On February 27, 2002, the system of organ allocation

Transcription

1 Persistent Ascites and Low Serum Sodium Identify Patients With Cirrhosis and Low MELD Scores Who Are at High Risk for Early Death Douglas M. Heuman, 1,3 Souheil G. Abou-assi, 1,3 Adil Habib, 1,3 Leslie M. Williams, 1 R. Todd Stravitz, 1 Arun J. Sanyal, 1 Robert A. Fisher, 2 and Anastasios A. Mihas 1,3 Despite the adoption of sickest first liver transplantation, pretransplant death remains common, and many early deaths occur despite initially low Model for End-stage Liver Disease (MELD) scores. From , we studied 507 cirrhotic United States veterans referred for consideration of liver transplantation to identify additional predictors of early mortality. Most of the patients were male (98%) with cirrhosis caused by hepatitis C and/or alcohol (88%). Data for 296 patients referred prior to February 27, 2002 (training group), were analyzed; findings were validated in 211 patients referred subsequently (validation group). In the training group, 61 patients (21%) died within 180 days without transplantation; their median initial MELD score was 21. MELD score, persistent ascites, and low serum sodium (<135 meq/l) were independent predictors of early mortality. In patients with a MELD score of less than 21, only low serum sodium and persistent ascites were independent predictors of mortality; for MELD scores above 21, only MELD was independently predictive. Prognostic significance of persistent ascites and low serum sodium for low MELD score patients was confirmed in the validation group. Risk varied continuously with worsening hyponatremia. Modifying MELD, by including points for persistent ascites and low serum sodium, improved prediction of early pretransplant mortality in low MELD score patients. In conclusion, persistent ascites and low serum sodium identify patients with cirrhosis with high mortality risk despite low MELD scores. Ascites, hyponatremia, and other findings indicative of hemodynamic decompensation merit further prospective study as prognostic indicators in patients awaiting liver transplantation, and should be considered in setting minimal listing criteria. Supplementary material for this article can be found on the HEPATOLOGY website ( html). (HEPATOLOGY 2004;40: ) Abbreviations: MELD, Model for End stage Liver Disease; INR, international normalized ratio; D.V.A., U.S. Department of Veterans Affairs; CTP, Child- Turcotte-Pugh; MELD-AS, MELD ascites sodium. From the Departments of 1 Medicine and 2 Surgery, Virginia Commonwealth University School of Medicine, Richmond, VA; and the 3 McGuire Department of Veterans Affairs Medical Center, Richmond, VA. Received December 14, 2003; accepted June 18, Address reprint requests to: Douglas M. Heuman, M.D., Medical Director of Hepatology and Liver Transplantation, GI Section (111-N), McGuire DVAMC, 1201 Broad Rock Blvd., Richmond VA douglas.heuman@med. va.gov; fax: Copyright 2004 by the American Association for the Study of Liver Diseases. Published online in Wiley InterScience ( DOI /hep On February 27, 2002, the system of organ allocation for liver transplantation in the United States changed to a sickest first approach, with priority based on a Model for End stage Liver Disease (MELD). 1,2 The MELD score, initially developed to predict mortality in patients with cirrhosis undergoing transjugular intrahepatic portosystemic shunt procedures, 3 has been shown to be a good predictor of short-term cirrhotic mortality in other settings. 4 7 MELD is particularly attractive as a basis for organ allocation because its components serum creatinine, bilirubin, and international normalized ratio (INR) are objective measurements not readily subject to bias or manipulation. 8 Although risk of death clearly is increased at high MELD scores, much of the early mortality in patients with cirrhosis still occurs in patients with low initial MELD. A rapid rise in MELD appears more predictive of impending death than absolute MELD score. 9,10 Better predictors are needed to identify patients with cirrhosis and low MELD scores who are at risk of rapid clinical deterioration. In particular, markers of advanced stages of cirrhotic hemodynamic derangement, such as persistent or refractory ascites, hy- 802

2 HEPATOLOGY, Vol. 40, No. 4, 2004 HEUMAN ET AL. 803 ponatremia, reduced free water clearance, increased plasma renin and norepinephrine, or low systemic arterial pressure may be important harbingers of hepatorenal failure and death In the current study, we analyzed clinical data prospectively recorded from patients with cirrhosis referred for consideration of liver transplantation within the U.S. Department of Veterans Affairs (D.V.A.) Health System prior to February 27, In this population, which consisted predominantly of men with noncholestatic liver disease, we found that half of all deaths from cirrhosis within 180 days occurred in patients with initial MELD scores of less than 21. Persistent ascites (including hydrothorax) and low serum sodium were strong independent predictors of 180-day mortality, especially in patients with cirrhosis and MELD scores below 21. This finding has important implications for sickest first liver transplantation, both in setting minimal listing criteria and in assigning priority for organ allocation. Study Design and Methods Data Collection. In the D.V.A health care system, liver transplantation is managed on a national level. Physicians at local D.V.A. facilities are responsible for identifying patients with cirrhosis in need of transplantation and for preparing and submitting an initial clinical evaluation (i.e., referral package). Physicians on the D.V.A. s national liver transplant committee then review each case. Once the application is approved, the patient is assigned to one of four regional D.V.A. transplant centers, where final on-site evaluation, listing, and transplantation take place. Between January 1997 and July 31, 2003, we reviewed 507 applications for orthotopic liver transplantation for adults with cirrhosis which contained the necessary basic information required to assess diagnosis and severity of liver disease, including etiological tests, clinical history of complications of cirrhosis, current status of ascites and encephalopathy, liver imaging studies, and basic laboratory studies (i.e., complete blood count, electrolyte and liver panels, and coagulation studies). Additional patients were excluded if referred for fulminant hepatic failure, if diagnosis of cirrhosis was suspect or absent, if the patient had undergone a prior transplant, or if insufficient data (concurrent within 30 days) were available to determine MELD and Child-Turcotte-Pugh (CTP) scores. All patients included in the analysis were according to their referring physicians and psychosocial evaluators abstinent from significant alcohol or drug abuse for at least 6 months (a precondition for transplantation in D.V.A. programs); negative screening tests for drugs and alcohol were also required. For purposes of assessing duration of survival, the initial date (time zero) was chosen to be the date of the most recent liver function test results available at the time of referral. Outcome (date of death or transplantation) was recorded and confirmed via: (1) D.V.A. computerized medical records, (2) reports transmitted to the D.V.A. central office (updated through July 1, 2003), and (3) searches of the Social Security Death Index (updated in August 2003). Factors analyzed included age, sex, race (when available); etiology of liver disease (alcohol, hepatitis C, hepatitis B, cholestatic disorders, other), presence of hepatocellular carcinoma (established or strongly suspected based on findings of mass or alpha fetoprotein 400), clinical complications of liver disease assessed semiquantitatively (severity of ascites and encephalopathy using the standard ordinal scale of the CTP score) or qualitatively (prior history of encephalopathy, ascites, spontaneous bacterial peritonitis, variceal hemorrhage, or therapeutic portosystemic shunt), and laboratory data (hemoglobin, erythrocyte indices, white blood cell count, platelet count, albumin, bilirubin, aspartate aminotransferase, alanine aminotransferase, aspartate aminotransferase/alanine aminotransferase ratio, alkaline phosphatase, serum sodium, creatinine, prothrombin time, and INR). To be classified as persistent, ascites (or hydrothorax) had to be of at least moderate severity despite diuretic treatment and confirmed by current imaging studies or recent paracentesis or thoracentesis. Further identification of the subset of patients with truly refractory ascites 18 was not attempted, because many referral packages lacked information regarding diuretic history and dietary salt restriction. These studies conformed to the ethical guidelines of the 1975 Helsinki declaration. Data collection and analysis were undertaken in compliance with federal regulations and with the approval and supervision of the McGuire Research Institutional Review Board. Statistical Analysis. Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) release 10.1 for Windows (SPSS, Inc., Chicago, IL). Analyses employed a training group consisting of 296 patients referred on or before February 27, 2002; findings were then validated in 211 patients referred subsequent to February 27, 2002 (the date of implementation of MELD-based organ allocation for liver transplantation in the United States). Missing data were excluded casewise. Predictors of death at 180 days were initially evaluated by univariate ANOVA (parametric) and Pearson chi square (nonparametric). Multivariable analysis employed binary logistic regression, performed incrementally with the forward stepwise approach. In

3 804 HEUMAN ET AL. HEPATOLOGY, October 2004 these analyses, nominal and ordinal variables were treated as categorical; CTP and MELD scores were analyzed as dimensional parameters. Probabilities for stepwise entry and removal were set at.05 and.10, respectively. Prognostic accuracy also was assessed using Cox proportionate hazard analysis with censoring at transplantation, and by means of receiver operating characteristic curves. Receiver operating characteristic curves of different prognostic indices were compared using the method described by Hanley and McNeil. 19 Results Age at the time of referral averaged (mean SD, range 28 74). The overwhelming majority was male (98%). Hepatitis C was present in 68%. Past alcohol use was felt to be a significant contributor to liver disease in 67% of referred patients, though only 23% had purely alcoholic cirrhosis. Hepatitis B surface antigenemia was present in 4%. Cholestatic liver diseases (primary biliary cirrhosis, primary sclerosing cholangitis, sarcoidosis) accounted for only 2.4% of referrals. Other etiologies, present in 9% of patients, included cryptogenic, nonalcoholic steatohepatitis, hereditary hemochromatosis, autoimmune hepatitis, alpha-1-antitrypsin deficiency, and sarcoidosis. Hepatocellular carcinoma was present at the time of referral in 12%. A significant increase in the mean age at referral (from in the training group to in the validation group) may reflect the aging of the D.V.A. population and in particular of the large cohort of veterans infected with hepatitis C. In all other respects, the training and validation cohorts were comparable. Population characteristics are summarized in Supplementary Table 1. Cirrhosis at the time of referral was relatively advanced, with a mean SD CTP score of ; 49% were CTP class C, 40% were class B, and 11% were class A. MELD score at referral was MELD scores (rounded to the nearest whole number) were 10 or less at the time of referral in 16%, in 59%, in 15%, and 25 or greater in 11%. Thirty-eight percent had persistent ascites and/or hydrothorax documented at the time of referral. Low serum sodium (less than 135 meq/l) was noted in 31%, and hyponatremia (less than 130 meq/l) in 11%. Severe recurrent or refractory encephalopathy was present in 15%. Although training and validation groups were comparable in most respects, patients referred after February 27, 2002, were slightly sicker, with a higher proportion of severe encephalopathy or low serum sodium and higher mean values of creatinine and prothrombin time. Clinical and laboratory indicators of severity of liver disease in training and validation groups are shown in detail in Supplementary Table 2. Fig. 1. Distribution of MELD scores among veterans with cirrhosis referred for transplantation prior to February 27, 2002 (training group). (A) Patients who survived more than 180 days without transplantation. (B) Patients who died within 180 days without undergoing transplantation. Among patients who died, mean MELD score at referral (dotted line) was 21.04; 29 of 61 early deaths occurred in patients with initial MELD scores below 21. Normal distribution is also shown for comparison (curved lines). MELD, Model for End stage Liver Disease. Overall life-table mortality from cirrhosis (censored at transplantation) at 90 days, 180 days, and 1 year from date of referral was 12%, 20%, and 39%, respectively. Only 18 patients (4%) were transplanted within 180 days. Pretransplant death thus was four to five times more likely than transplantation during the first 180 days. The proportion of patients transplanted within 180 days increased significantly from 2% in the training subgroup to 7% in the validation group (P.006), probably as a consequence of the implementation of sickest first transplant allocation after February 27, The distribution of MELD scores in 61 patients in the training group who died of cirrhosis within 180 days of referral, compared with 229 who survived for more than 180 days without transplantation, is shown in Fig. 1. Mean MELD scores of deaths versus survivors were 23 8 and 14 4, respectively (mean SD, P.001). The

4 HEPATOLOGY, Vol. 40, No. 4, 2004 HEUMAN ET AL. 805 Table 1. Results of Multivariate Analysis (Binary Logistic Regression), Training Group Predictors Included in Multivariate Logistic Regression Analysis Significance If Removed Continuous Albumin.341 Bilirubin.35 INR.542 Creatinine.068 AST/ALT ratio.866 MELD score.001 CTP score.244 Sodium.289 Categorical History of ascites (CTP ascites score 2 or 3).828 Persistent ascites (CTP ascites score 3).004 History of encephalopathy.828 Serum sodium 135 meq/l.003 History of spontaneous bacterial peritonitis.493 Regression Coefficient (B) SE Significance (P) Exp(B) (Likelihood Ratio) 95% Confidence Interval (Likelihood Ratio) Low High Weighting of independent predictors by logistic regression* MELD score Serum sodium 135 meq/l Persistent ascites Independent predictive value of MELD, sodium, and persistent ascites in patients with MELD scores 21 MELD score Serum sodium 135 meq/l NS Persistent ascites NS Independent predictive value of MELD, sodium 135 meq/l, and persistent ascites in patients with MELD scores 21 MELD score NS Serum sodium 135 meq/l Persistent ascites Abbreviations: AST, aspartate aminotransferase; ALT, alanine aminotransferase; NS, not significant. *Optimized logistic model incorporating MELD, ascites, and low sodium ( MAS) normalized to B (MELD) 1. MELD-AS MELD 4.53 (sodium 135 meq/l [0.1]) 4.46 (persistent ascites (0.1]). median initial MELD score of patients dying within 180 days was 21.04; of the 61 deaths, 29 occurred in the 250 patients with an initial MELD score of less than 21. According to univariate analysis, overall predictors of mortality significant at P.05 included history of spontaneous bacterial peritonitis, history of ascites, persistent ascites or hydrothorax, history of encephalopathy, low serum sodium, and a number of laboratory parameters (white blood cell count, hemoglobin count, prothrombin time, INR, bilirubin, albumin, creatinine, sodium, and aspartate aminotransferase/alanine aminotransferase ratio). Both MELD and CTP scores also were strongly associated with early mortality (P.001 for both). In patients who had a MELD score above 21, only MELD score was significantly associated with 180-day mortality. No other demographic, clinical, or laboratory parameter was found to distinguish cirrhotic deaths from survivors in this high MELD score group. In contrast, in patients who had a MELD score lower than 21, a number of other factors were associated with 180-day mortality, including low serum sodium, anemia, leukocytosis, persistent ascites, elevated aspartate aminotransferase/alanine aminotransferase ratio, creatinine, sodium, and CTP score. Detailed results of univariate analysis are shown in Supplementary Table 3. In a multivariate analysis (Table 1), we elected to exclude the hematological parameters because of difficulty controlling for transient effects of acute hemorrhage, transfusion, and infection. Prothrombin time was omitted as redundant with INR. When the remaining univariate predictors significant at P.05 were subjected to binary logistic regression, only MELD score, persistent ascites, and low serum sodium ( 135 meq/l) emerged as independent predictors. Of these three factors, only MELD was an independent predictor in patients with MELD scores above 21. Conversely, multivariable analy-

5 806 HEUMAN ET AL. HEPATOLOGY, October 2004 Fig. 2. Kaplan-Meier fractional survival curves for the training group (296 patients referred prior to February 27, 2002), censored at transplantation. The complementary effect of persistent ascites and low serum sodium on 180-day survival is apparent, especially in patients with MELD scores below 21. Data stratified by the presence or absence of (A C) persistent ascites, (D F) low serum sodium, or (G I) combined ascites and sodium of less than 135 meq/l. Analyses in the left-hand column (A, D, G) include all patients; middle column (B, E, H), patients with MELD scores below 21; and right-hand column (C, F, I), patients with MELD scores above 21. MELD, Model for End stage Liver Disease; Asc, ascites. sis of these three factors in patients with MELD scores below 21 found only persistent ascites and low serum sodium to be predictive; MELD was not an independent predictor of outcome in the low MELD score patients once persistent ascites and serum sodium were taken into account. Serum sodium was an independent predictor of survival for patients with MELD scores below 21 whether analyzed as a continuous variable or as a categorical variable using a cutoff of 135 meq/l. The effects of persistent ascites and low serum sodium on 180-day survival in the training group are shown in Fig. 2 and Table 2. Low serum sodium was present in 51 patients, of whom 17 died (33.3%); in patients with initially normal sodium, mortality at 180 days was only 5.8%. A trend was noted toward greater 180-day mortality in patients with hyponatremia (sodium 130 meq/l) compared with those with mildly low serum sodium ( meq/l). Persistent ascites was present in 74 patients, with 19 deaths (25.7%), versus mortality of only 5.8% in patients without persistent ascites. Of 63 patients exhibiting either low serum sodium or persistent ascites (but not both), 11 died within 180 days (17.5%). In the 29 patients exhibiting both persistent ascites and low serum sodium, there were 12 deaths (41.4%). Either persistent ascites or low serum sodium was present at referral in 82% of low MELD patients who died within 180 days, but only 32% of those who survived. In patients with a MELD score above 21, no effect of persistent ascites or low serum sodium on survival could be identified, but most patients in this group (37 of 44, 84%) had one or both of these findings. Multivariable analysis using binary logistic regression was repeated to determine the relative predictive weight of MELD, persistent ascites, and low serum sodium for all patients in the training group, as shown in Table 3. With weighting coefficient of MELD normalized to a value of 1, logistic regression assigned nearly identical coefficients of 4.46 and 4.53 to persistent ascites and low serum so- Table 2. Likelihood of Death in Cirrhotic Patients (Censored at Transplantation) as a Function of Persistent Ascites and/or Low Serum Sodium (<135 meq/l) as Assessed via Cox Proportional Hazards Model Training Group Validation Group Likelihood ratio (LR) 95% CI for LR 95% CI for LR Lower Upper Significance (P) Likelihood ratio (LR) Lower Upper Significance (P) All patients 0 vs vs vs MELD 21 0 vs vs vs MELD 21 0 vs vs vs NOTE. 0, Ascites absent, normal sodium; 1, either persistent ascites or low sodium; 2, both persistent ascites and low sodium.

6 HEPATOLOGY, Vol. 40, No. 4, 2004 HEUMAN ET AL. 807 Table 3. Concordance Statistics of Different Indices (MELD Score, CTP Score, Modified MELD-AS Score) as Predictors of 180-day Cirrhotic Patient Mortality, Derived From Analysis of Receiver-Operating Characteristic Curves Training Group Validation Group Combined Groups All MELD Died (n) Survived (n) CTP score MELD score MELD-AS score * * MELD 21 Died (n) Survived (n) CTP score MELD score MELD-AS score * * MELD 21 Died (n) Survived (n) CTP score MELD score MELD-AS score * *.001). The combination of persistent ascites and serum sodium level of less than 135 meq/l was noted in 22 patients, of whom 10 died within 180 days (45.5%); in 88 patients with neither persistent ascites nor low serum sodium, mortality was only 4.4%. In patients with high MELD scores ( 21), there was no significant association between mortality and the presence of persistent ascites or low serum sodium. However, an association could easily have been missed, because either low sodium or persistent ascites was present in 17 of the 19 high MELD score patients, and both were present in 9 patients. The predictive value of serum sodium as a continuous prognostic variable was further characterized in analyses of the combined training and validation cohorts (see Supplementary Fig. 1). Through receiver-operating characteristic curve analysis including all patients, serum sodium as a predictor of 180-day mortality was associated with a concordance statistic (area under the receiver operating characteristic curve) of (mean SEM; 95% CI ). Comparable concordance statis- *P.05, MELD-AS versus CTP score. P.05, MELD-AS versus MELD score. P.05, MELD versus CTP score. dium ( 135 meq/l), respectively. This indicates that optimal identification of patients at risk of 180-day pretransplant death can be obtained by adding 4.46 points to the MELD score if persistent ascites is present and 4.53 points if sodium is less than 135 meq/l. We propose that the modified score derived in this manner be termed the MELD-AS score (an acronym for MELD ascites sodium). The importance of persistent ascites and low serum sodium as predictors of early mortality in patients with cirrhosis was confirmed in patients referred after February 27, 2002 (validation group), as shown in Fig. 3. Of 211 patients with cirrhosis, 180-day transplant-free follow-up data were available for 154, 26 of whom died within this period. Among patients with a MELD score above 21 at referral, pretransplant mortality was 10 of 19 (52.6%), compared with 16 of 135 (11.9%) among patients with a MELD score below 21. In the low MELD score patients, serum sodium levels of less than 135 meq/l in 45 patients was associated with 180-day mortality of 24.4% versus 5.7% in the 88 patients without hyponatremia (P.002); among patients with serum sodium levels of less than 130 meq/l, mortality was 50% (6 of 12; P.001 compared with a serum sodium level above 130 meq/l). Similarly, persistent ascites in 42 patients was associated with 28.6% mortality, compared with 4.5% in the 93 patients in whom ascites was absent or controlled (P Fig. 3. Validation of persistent ascites and hyponatremia as predictors of 180-day mortality in cirrhotic veterans, using data derived from patients referred for consideration of transplantation on or after February 27, 2002 (validation group). Kaplan-Meier fractional survival, censored at transplantation. The adverse effect of persistent ascites and low serum sodium on 180-day survival, especially in patients with MELD scores below 21, is confirmed. Data stratified by the presence or absence of (A C) persistent ascites, (D F) low serum sodium, or (G I) combined ascites and sodium of less than 135 meq/l. Analyses in left-hand column (A, D, G) include all patients; middle column (B, E, H), patients with MELD scores below 21; and right-hand column (C, F, I), patients with MELD scores above 21. MELD, Model for End stage Liver Disease; Asc, ascites.

7 808 HEUMAN ET AL. HEPATOLOGY, October 2004 Fig. 4. Receiver-operating characteristic curves comparing MELD score, CTP score, and MELD-AS score (logistic model incorporating MELD, persistent ascites, and low serum sodium 135 meq/l), as predictors of 180-day pretransplant mortality. Data shown are for all patients in combined training and validation groups with more than 180-day transplant-free followup. (A) All patients (n 438, 87 deaths). (B) Patients with initial MELD scores below 21 (n 375, 45 deaths). (C) Patients with initial MELD scores above 21 (n 63, 42 deaths). MELD-AS significantly outperformed MELD and CTP scores, both overall and for patients with MELD scores below 21; for MELD scores above 21, the MELD and MELD-AS scores were equivalent. For statistical comparisons, see Table 3. Abbreviations: MELD-AS, MELD ascites sodium; MELD, Model for End stage Liver Disease; CTP, Child-Turcotte-Pugh. tics for bilirubin, creatinine, albumin, and INR were , , , and , respectively (data not shown). A serum sodium level of less than 135 meq/l was 59% specific and 77% sensitive for death within 180 days; a cut-off of 130 meq/l was 94% specific and 27% sensitive. Serum sodium was especially predictive in patients with MELD scores below 21. The concordance statistic for serum sodium as a predictor of 180-day mortality in this low MELD score group was , whereas concordance statistics for bilirubin, creatinine, albumin, and INR were only , , , and , respectively. In contrast, sodium was not predictive in patients with MELD scores above 21. Through Cox regression, mortality of patients with normal serum sodium levels was significantly less than that of patients with mildly low serum sodium levels of meq/l (P.001; hazard ratio 2.21; 95% CI ) or more severe hyponatremia ( 130 meq/l) (P.001; hazard ratio 4.86; 95% CI ). The predictive accuracy of the MELD-AS logistic regression model, incorporating MELD, persistent ascites, and low serum sodium (see Table 1), was validated in the cohort of patients referred on or after February 27, Results of receiver operating characteristic curve analysis are shown in Fig. 4 and Table 3. In both the training and validation groups as a whole, the MELD-AS model significantly outperformed the MELD score, and for the combined populations it also outperformed the CTP score as a predictor of 180-day mortality. The advantage of MELD-AS over MELD was most apparent in patients with MELD scores below 21. In high MELD score patients, MELD and the modified MELD-AS model were comparable predictors. Of note, concordance statistics for the MELD and CTP scores as predictors of 180-day mortality in the training group (0.83 and 0.79, respectively) were similar to those reported by Wiesner et al. 20 for these scores as predictors of 90-day mortality on the United Network for Organ Sharing national transplant list in 2001 (0.83 and 0.76, respectively). The lower accuracy of MELD and MELD-AS scores as predictors of pretransplant survival in the validation group compared with the training group may have resulted from more effective identification and transplantation of high-risk patients with adoption of MELD-based organ allocation after February 27, Discussion The adoption of sickest first organ allocation was a significant step forward in liver transplantation, 20 but the current MELD-based system is not without its problems. The MELD score is disproportionately increased by cholestasis with hyperbilirubinemia or intrinsic renal disease, and thus conveys a relative transplant advantage to patients who have these conditions. MELD-based organ allocation specifically excludes any role for clinical judgment and does not allow any consideration of disability or quality of life. Most importantly, MELD-based organ allocation as currently structured does not consistently allow for consideration of other laboratory or clinical findings that may affect prognosis, except in patients with hepatocellular carcinoma or hepatopulmonary syndrome. The potential limitations of MELD have been appreciated even by many of its strongest proponents, who have acknowledged that addition of other variables might refine and improve the accuracy of these risk models. 1 In the current study, we have demonstrated that hyponatremia and persistent ascites are predictors of early mortality independently of MELD, and are especially important in patients with MELD scores below 21. A substantial proportion of patients in the United States with MELD scores in this range receive transplants, and a substantial proportion of early mortality from cirrhosis occurs in patients with recently low MELD scores. In a recent report of 2,745 adults who received transplants in the United States between February 28, 2002 and October 31, 2002, it was noted that over 60% had MELD scores of less than 19.5 at transplantation. 21 Both in our patients and in the United Network for Organ Sharing transplant waiting list in the immediate pre-meld era, 22 median initial MELD scores in patients dying within days averaged a MELD score of 21 or less.

8 HEPATOLOGY, Vol. 40, No. 4, 2004 HEUMAN ET AL. 809 Although this study is the first to demonstrate that hyponatremia and persistent ascites have short-term prognostic significance independent of MELD score, the association of persistent ascites and hyponatremia with mortality in cirrhosis is neither new nor surprising. Numerous previous studies have shown that the severity of cirrhotic sodium and water retention is a major determinant of disease severity and prognosis. 12,14 17,23,24 Ascites and hyponatremia are manifestations of the generalized hemodynamic derangement of cirrhosis, with its low peripheral vascular resistance, reduced effective circulating volume, central overproduction of antidiuretic hormone, elevated renin, angiotensin, and norepinephrine, reduced glomerular filtration, and marked renal salt and water retention. 25,26 Hyponatremia is one of the factors limiting diuretic dosage in ascites treatment and therefore may be a factor in rendering ascites refractory. Refractory ascites has long been known to have a poor prognosis, with 1-year mortality of 50% 90% The prognostic value of persistent ascites and hyponatremia probably reflects the fact that these features characterize patients with relatively advanced hemodynamic derangement who are at high risk of progressing to type 1 hepatorenal syndrome. 30,31 High MELD scores typically occur in patients with established hepatorenal syndrome, and rapidly rising MELD scores 9,10 largely identify patients with incipient type 1 hepatorenal syndrome, often precipitated by infection. Such patients have high pretransplant mortality, complicated hospital courses with increased peritransplant costs, 32 and poorer patient and allograft survival. 21,33,34 Transplantation prior to terminal decompensation and renal failure would be expected to improve these outcomes. The current study was performed in a relatively homogeneous population of middle-aged males with noncholestatic cirrhosis and with at least 6 months documented abstinence from alcohol and illicit drugs. The general applicability of the findings needs to be confirmed further by prospective evaluation in other populations with larger proportions of female patients and a more diverse spectrum of liver diseases. However, the principal predictors that have been identified hyponatremia and persistent ascites are manifestations of the underlying hemodynamic derangement common to all forms of advanced cirrhosis, and it is therefore likely that similar results will be found in a broad spectrum of patients with cirrhosis. Persistent ascites and hyponatremia, while objective parameters that are readily documented, are relatively crude indicators of the underlying circulatory derangement of cirrhosis. Severity of ascites and hyponatremia varies with the aggressiveness of diuretic treatment as well as salt and water intake, and both might be subject to manipulation in the transplant setting. More objective and reproducible measures that quantify the severity of circulatory dysfunction might permit more precise identification of patients with cirrhosis and low MELD score with high short-term mortality risk. Patients who have truly refractory ascites as defined by the rigorous criteria of the International Ascites Club clearly would represent a higher-risk subpopulation. However, we found that even mild degrees of low serum sodium in patients with controlled ascites were associated with increased short-term mortality. The data of Fernandez-Esparrach et al. 17 indicate that, for patients hospitalized with ascites, impaired renal free water excretion in response to a water challenge and low mean arterial blood pressure were predictive of survival, independent of serum creatinine and CTP score. Further exploration of these and other markers of cirrhotic circulatory dysfunction is warranted for use as prognostic indicators in the transplant setting. An immediate application of our findings may be in the area of minimal listing criteria. It has been suggested that patients with low MELD scores have an excellent prognosis and need not be listed for transplantation. Our data indicate that this view requires further refinement. The subset of low MELD score patients with low serum sodium and persistent ascites has substantial early mortality. In the presence of both of these findings, the risk of pretransplant death within 180 days exceeds 40%. Such patients clearly would benefit from earlier transplantation regardless of MELD score; under the current MELDbased organ allocation criteria, they may merit strong consideration for living donor liver transplantation. Conversely, patients with MELD scores below 21 who exhibited neither hyponatremia nor persistent ascites had a very good prognosis (6-month mortality 5%) and may represent a population in whom transplantation can safely be deferred. In summary, we have demonstrated that persistent ascites and low serum sodium are important independent predictors of early pretransplant mortality, especially for patients with MELD scores below 21. Modification of MELD by inclusion of points for persistent ascites and low serum sodium as dictated by logistic regression may improve predictive accuracy, especially at lower MELD scores. Our data strongly support the conclusion that patients with cirrhosis and MELD scores below 21 who exhibit persistent ascites and low serum sodium merit expedited consideration for liver transplantation under a sickest first model. Conversely, the absence of persistent ascites or hyponatremia in patients with cirrhosis and MELD scores below 21 is indicative of a favorable shortterm prognosis, and identifies a population in whom transplantation may safely be deferred. Serum sodium,

9 810 HEUMAN ET AL. HEPATOLOGY, October 2004 persistent ascites, and other markers of the hemodynamic derangement of advanced cirrhosis should be included in future large-scale data collections, such as the United Network for Organ Sharing transplant database, so that their prognostic significance can be confirmed prospectively in a broader context. Acknowledgment: The authors gratefully acknowledge the clerical assistance of Chauvonna Taylor and Cecile Rock. Brenda Salvas, transplant manager for the Department of Veterans Affairs, provided invaluable assistance in tracking outcomes of transplant referrals. We thank Denise Tripp and Patricia Paquette of the United Network for Organ Sharing (Richmond, VA) for sharing data from the Organ Procurement and Transportation Network national database. References 1. Freeman RB Jr, Wiesner RH, Harper A, McDiarmid SV, Lake J, Edwards E, et al. The new liver allocation system: moving toward evidence-based transplantation policy. Liver Transpl 2002;8: Wiesner RH, McDiarmid SV, Kamath PS, Edwards EB, Malinchoc M, Kremers WK, et al. MELD and PELD: application of survival models to liver allocation. Liver Transpl 2001;7: Malinchoc M, Kamath PS, Gordon FD, Peine CJ, Rank J, ter Borg PC. A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts. HEPATOLOGY 2000;31: Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM, Kosberg CL, et al. A model to predict survival in patients with end-stage liver disease. HEPATOLOGY 2001;33: Botta F, Giannini E, Romagnoli P, Fasoli A, Malfatti F, Chiarbonello B, et al. MELD scoring system is useful for predicting prognosis in patients with liver cirrhosis and is correlated with residual liver function: a European study. Gut 2003;52: Sheth M, Riggs M, Patel T. Utility of the Mayo End-Stage Liver Disease (MELD) score in assessing prognosis of patients with alcoholic hepatitis. BMC Gastroenterol 2002;2:2. 7. Freeman RB Jr. Around the world with the model for end- stage liver disease. Liver Transpl 2003;9: Forman LM, Lucey MR. Predicting the prognosis of chronic liver disease: an evolution from child to MELD. Mayo End-stage Liver Disease. HEPA- TOLOGY 2001;33: Merion RM, Wolfe RA, Dykstra DM, Leichtman AB, Gillespie B, Held PJ. Longitudinal assessment of mortality risk among candidates for liver transplantation. Liver Transpl 2003;9: Kamath PS, Kim WR. Is the change in MELD score a better indicator of mortality than baseline MELD score? Liver Transpl 2003;9: Shear L, Kleinerman J, Gabuzda GJ. Renal failure in patients with cirrhosis of the liver. I. Clinical and pathological characteristics. Am J Med 1965; 39: Arroyo V, Rodes J, Gutierrez-Lizarraga MA, Revert L. Prognostic value of spontaneous hyponatremia in cirrhosis with ascites. Am J Dig Dis 1976; 21: Arroyo V, Bosch J, Gaya-Beltran J, Kravetz D, Estrada L, Rivera F, et al. Plasma renin activity and urinary sodium excretion as prognostic indicators in nonazotemic cirrhosis with ascites. Ann Intern Med 1981;94: Llach J, Gines P, Arroyo V, Rimola A, Tito L, Badalamenti S, et al. Prognostic value of arterial pressure, endogenous vasoactive systems, and renal function in cirrhotic patients admitted to the hospital for the treatment of ascites. Gastroenterology 1988;94: Gines P, Quintero E, Arroyo V, Teres J, Bruguera M, Rimola A, et al. Compensated cirrhosis: natural history and prognostic factors. HEPATOL- OGY 1987;7: Cosby RL, Yee B, Schrier RW. New classification with prognostic value in cirrhotic patients. Miner Electrolyte Metab 1989;15: Fernandez-Esparrach G, Sanchez-Fueyo A, Gines P, Uriz J, Quinto L, Ventura PJ, et al. A prognostic model for predicting survival in cirrhosis with ascites. J Hepatol 2001;34: Arroyo V, Gines P, Gerbes AL, Dudley FJ, Gentilini P, Laffi G, et al. Definition and diagnostic criteria of refractory ascites and hepatorenal syndrome in cirrhosis. International Ascites Club. HEPATOLOGY 1996;23: Hanley JA, McNeil BJ. A method of comparing the areas under receiver operating characteristic curves derived from the same cases. Radiology 1983;148: Wiesner R, Edwards E, Freeman R, Harper A, Kim R, Kamath P, et al. Model for end-stage liver disease (MELD) and allocation of donor livers. Gastroenterology 2003;124: Wiesner RH, Edwards E, Freeman R, Harper A. Does the Model for End stage Liver Disease (MELD) predict post liver transplant graft survival. HEPATOLOGY 2003;38:371A. 22. Heuman DM, Mihas A. Utility of the MELD score for assessing 3-month survival in patients with liver cirrhosis: one more positive answer. Gastroenterology 2003;125: Shear L, Kleinerman J, Gabuzda GJ. Renal failure in patients with cirrhosis of the liver. I. Clinical and pathological characteristics. Am J Med 1965; 39: Arroyo V, Bosch J, Gaya-Beltran J, Kravetz D, Estrada L, Rivera F, et al. Plasma renin activity and urinary sodium excretion as prognostic indicators in nonazotemic cirrhosis with ascites. Ann Intern Med 1981;94: Arroyo V, Bernardi M, Epstein M, Henriksen JH, Schrier RW, Rodes J. Pathophysiology of ascites and functional renal failure in cirrhosis. J Hepatol 1988;6: Arroyo V, Colmenero J. Ascites and hepatorenal syndrome in cirrhosis: pathophysiological basis of therapy and current management. J Hepatol 2003;38(Suppl 1):S69 S Sanyal AJ, Genning C, Reddy KR, Wong F, Kowdley KV, Benner K, et al. The North American Study for the Treatment of Refractory Ascites. Gastroenterology 2003;124: Guardiola J, Xiol X, Escriba JM, Castellvi JM, Castellote J, Baliellas C, et al. Prognosis assessment of cirrhotic patients with refractory ascites treated with a peritoneovenous shunt. Am J Gastroenterol 1995;90: Gines P, Arroyo V, Vargas V, Planas R, Casafont F, Panes J, et al. Paracentesis with intravenous infusion of albumin as compared with peritoneovenous shunting in cirrhosis with refractory ascites. N Engl J Med 1991; 325: Porcel A, Diaz F, Rendon P, Macias M, Martin-Herrera L, Giron-Gonzalez JA. Dilutional hyponatremia in patients with cirrhosis and ascites. Arch Intern Med 2002;162: Gines A, Escorsell A, Gines P, Salo J, Jimenez W, Inglada L, et al. Incidence, predictive factors, and prognosis of the hepatorenal syndrome in cirrhosis with ascites. Gastroenterology 1993;105: Shetty K, Sonnad S, Olthoff KM, Shaked A, Reddy RK. Impact of the MELD score on resource utilization following liver transplantation. HEPA- TOLOGY 2003;38:367A. 33. Onaca NN, Levy MF, Sanchez EQ, Chinnakotla S, Fasola CG, Thomas MJ, et al. A correlation between the pretransplantation MELD score and mortality in the first two years after liver transplantation. Liver Transpl 2003;9: Saab S, Wang V, Ibrahim AB, Durazo F, Han S, Farmer DG, et al. MELD score predicts 1-year patient survival post-orthotopic liver transplantation. Liver Transpl 2003;9: