Pre-transplant MELD and sodium MELD scores are poor predictors of graft failure and mortality after liver transplantation

Transcription

1 Hepatol Int (2011) 5: DOI /s z ORIGINAL ARTICLE Pre-transplant MELD and sodium MELD scores are poor predictors of graft failure and mortality after liver transplantation Jacek B. Cywinski Edward J. Mascha Jing You Daniel I. Sessler Leonardo Kapural Maged Argalious Brian M. Parker Received: 26 July 2010 / Accepted: 17 January 2011 / Published online: 17 February 2011 Ó Asian Pacific Association for the Study of the Liver 2011 Abstract Background Incorporating serum sodium concentrations into the model for end-stage liver disease (MELD) score may increase its sensitivity for identifying priority patients for orthotopic liver transplantation (OLT). We, therefore, evaluated and compared the ability of the sodium MELD and MELD scores to predict graft and patient survival after OLT. Methods The United Network for Organ Sharing (UNOS) registry includes all US adult OLTs performed between January 2000 and August For 15,156 patients who met inclusion criteria, MELD score was calculated; for 6,193 patients whose serum sodium concentrations was J. B. Cywinski (&) Departments of General Anesthesiology and Outcomes Research, Transplantation Center, Cleveland Clinic, Cleveland, OH, USA cywinsj@ccf.org E. J. Mascha J. You Departments of Quantitative Health Sciences and Outcomes Research, Cleveland Clinic, Cleveland, OH, USA D. I. Sessler Department of Outcomes Research, Cleveland Clinic, Cleveland, OH, USA L. Kapural Departments of Outcomes Research and Pain Management, Cleveland Clinic, Cleveland, OH, USA M. Argalious Department of General Anesthesiology, Cleveland Clinic, Cleveland, OH, USA B. M. Parker Department of General Anesthesiology and Transplantation Center, Cleveland Clinic, Cleveland, OH, USA between 120 and 135 meq/dl, immediately before OLT, sodium MELD score was calculated. The corresponding hazard ratios (HR) for MELD and sodium MELD on graft and patient survival were assessed using the Cox proportional hazards regression models. The concordance probability estimate (CPE) was used to evaluate predictive ability of each time-to-event model. Results MELD and sodium MELD scores were both significant predictors in univariable Cox regression models for graft failure [HR (95% CI) for every 10 units increase in the predictor: 1.10 (1.04, 1.17), P = 0.001, and 1.05 (1.00, 1.10), P = 0.03, respectively], and for mortality (1.14 (1.07, 1.21), P \ 0.001, and 1.07 (1.02, 1.12), P = 0.01, respectively), with CPE of Conclusion While MELD and sodium MELD were each significantly associated with survival after OLT, their predictive abilities were poor. The sodium MELD score does not improve prediction accuracy over the MELD score. Weak prediction may result from unaccounted variability in recipient and donor status, as well as surgical and postoperative factors. Keywords Hyponatremia Recipient Model end-stage liver disease Orthotopic liver transplantation Outcomes Introduction Since the introduction of the model for end-stage liver disease (MELD) score for organ allocation in 2002, both the orthotopic liver transplantation (OLT) waiting list mortality and waiting time have decreased [1]. The MELD score is based on objective parameters: serum creatinine, serum bilirubin, and the international normalized ratio. Many investigators have tried to improve MELD score

2 842 Hepatol Int (2011) 5: accuracy for predicting preoperative mortality by including additional potentially important factors [2, 3]. One of these efforts focused on the addition of the recipient serum sodium concentration to the model [2, 3], based on the theory that low-serum sodium concentrations may identify patients with advanced disease state. For example, Ruf et al. [4] found that by incorporating serum sodium concentration into the MELD score, renal impairment and/or circulatory dysfunction are identified at an earlier phase in patients with advanced cirrhosis. They, thus, concluded that inclusion of serum sodium in the MELD score better detects patients predisposed to poor outcome, thereby improving the ability of the MELD score alone to predict transplant waiting list mortality [4]. Although there is a general consensus that the MELD score is an excellent predictor of mortality for patients awaiting OLT, it remains debatable whether the pretransplant MELD score predicts postoperative outcomes [1, 2, 5, 6]. Some studies have found that the pre-transplant MELD score had a significant influence on survival during the first two postoperative years [7 10], while others question its value as a tool to predict both post-transplant outcome and survival [5, 11]. Interestingly, MELD score may have better sensitivity to predict post-transplant outcomes in patients with hepatitis C-induced end-stage liver disease. Onaca et al. [12] found that pre-transplant MELD score above 24 in patients with hepatitis C-induced liver disease was associated with significantly lower survival, questioning transplant suitability for these patients. Wang et al. [10] found in a small study that MELD score was an objective predictive system and more efficient than Child- Turcotte-Pugh (CTP) score to evaluate the risk of 3-month morbidity and short-term prognosis in patients with acuteon-chronic hepatitis B liver failure undergoing OLT. However, it still remains debated if one can improve MELD score ability to predict post-transplant outcome in all recipients regardless of etiology of end-stage liver disease (ESLD). If addition of recipient serum sodium concentration to MELD indeed better detects the sickest pretransplant patients than the MELD score alone, it seems likely to also better predict outcomes and mortality after OLT. In particular, recipient hyponatremia has been associated with an increased rate of neurologic disorders, infectious complications, and renal failure during the first month after transplantation and reduced 3-month survival [13]. Pre-transplant hyponatremia may even indicate multiorgan derangements which are not accounted for in the MELD scoring system, and thus increased risk for postoperative complications. Addition of serum sodium to the MELD score may, therefore, improve ability to predict postoperative outcomes. Consistent with this theory, Dawwas et al. [14] demonstrated in more than 5,000 patients that severe recipient hyponatremia (serum sodium \130 meq/dl) was associated with increased post-transplant mortality at 90 days. We hypothesized that the sodium MELD score, calculated according to the formula developed by Biggins et al. [2], will predict graft failure and mortality of the hyponatremic OLT recipients better than conventional MELD score. Methods The Cleveland Clinic Institutional Review Board approved this registry analysis. Data were obtained from the United Network for Organ Sharing (UNOS) database, and included patients who had OLT in the US between January 2000 and August 2008 [15]. Among 46,703 patients, immediate preoperative serum sodium concentrations were available for 24,297 and these patients were screened for inclusion. Only results from the initial transplant surgery were considered, leaving 22,393 patients. Patients who were under 18 years of age, lost follow-up within 6 months after the surgery, or with missing MELD scores or with both outcomes were also excluded. Thus, 15,156 patients were included for the analyses. The sodium MELD score was calculated according to the formula of Biggins et al. [2] in patients (n = 6,186) with pre-operative serum sodium concentrations between 120 and 135 meq/dl, as sodium MELD = MELD? (135 - serum sodium). Our outcomes were time to graft failure and time to allcause mortality. Graft failure time was defined as the time from transplant to the earliest listing of either failure of graft, re-transplant, or death from liver disease. For both outcomes, patients with follow-up time [3 years (about 5% of patients, 1% of events) were censored at 3 years, in order to reduce noise and potential bias in the survival analysis. Primary analyses We evaluated the association between each predictor (i.e., the MELD and the sodium MELD scores) and each outcome after the OLT surgery (i.e., graft failure and allcause mortality) for the subset of the patients whose serum sodium concentrations were between 120 and 135 meq/dl immediately before OLT. Each association was assessed using a Cox proportional hazards regression model [16]; hazard ratios (HR) with 95% confidence interval (CI) were estimated. The concordance probability estimate (CPE) [17], which ranges from 0.50 (association by chance) to 1.0 (perfect prediction), was used as a measure of predictive ability of each time-to-event model. Kaplan and Meier [18]

3 Hepatol Int (2011) 5: density functions and their associated equal-precision 95% CIs were estimated and plotted; the number of events was summarized at various post-transplant time points as well. Secondary analyses We further explored the relationships between the sodium MELD score and both graft failure and all-cause mortality. First, we visually assessed each relationship by plotting the logarithm of the hazard ratio for the predictor (MELD or sodium MELD) across the range of values of the predictor using the penalized splines [18] (P-splines) smoothing approach. For relationships appearing nonlinear, we then chose cut points using a searching algorithm for the maximum likelihood across the plausible range of cut points for each predictor. Based on these results, we fitted the predictor (MELD or sodium MELD score) as a continuous variable in a Cox proportional hazard regression model with an appropriate term (i.e., linear, piecewise, or polynomial) and reported the appropriate hazard ratio(s) (95% CI). Patient groups based on the cut points were compared on the outcomes. We also evaluated the abilities of both the MELD and the sodium MELD scores to predict an earlier (i.e., 30 days) graft failure or mortality event by Cox proportional hazards regressions. Patients who did not experience the event were censored at 30 days. Finally, we assessed, using Cox proportional hazard regression models, whether the relationships between each predictor (i.e., MELD and sodium MELD scores) and each outcome (i.e., graft failure and mortality) depend on other variables [including recipient age, gender, BMI, cause of ESLD (including HBV and HCV), donor age, gender, BMI, total cold ischemic time, and donor-risk index]. SAS software version 9.2 (SAS institute, Cary, NC, USA) and R statistical software version (The R Foundation Table 1 Recipient and donor baseline characteristics for the primary surgery in patients with pre-operative serum sodium concentration between 120 and 135 (meq/dl), N = 6,186, and in all patients, N = 15,156 Data presented as median (1st, 3rd quartiles) or percent of patients MELD model for end-stage liver disease, INR international normalized ratio a \1.5% b 5 6% c 11 16% of patients had missing values d Sodium MELD score = MELD? 1.59 (135- serum sodium) [2] Variables Patients with pre-op. serum sodium meq/dl (N = 6,186) All patients (N = 15,156) Recipient Age, years 54 (49, 59) 54 (48, 60) Gender, female (%) Body mass index (kg/m 2 ) 28 (25, 32) a 28 (25, 32) a Serum Sodium (meq/dl) 132 (130, 134) 137 (133, 139) Serum Creatinine (mg/dl) 1.2 (0.9, 1.8) 1.1 (0.8, 1.7) INR 1.7 (1.4, 2.2) 1.6 (1.3, 2.0) Bilirubin (mg/dl) 4.3 (2.3, 9.1) 3.4 (1.7, 7.7) MELD score 21.2 (16.1, 27.4) 18.9 (13.6, 26.1) Sodium MELD score d 27.0 (19.9, 35.3) N/A Ascites (%) Moderate 36 a 27 a Slight Absent Share Type (%) Local Regional National 7 8 Donor Age, years 43 (26, 55) 43 (26, 55) Gender, female (%) Body mass index (kg/m 2 ) 26 (23, 30) a 26 (23, 30) a Total cold ischemic time (hour) 7 (5, 9) c 7 (5, 9) c Risk index 1.4 (1.1, 1.7) c 1.4 (1.1, 1.7) c Ethnicity (%) White Black Hispanic Others 3 3 Cardiac arrest (%) 7 b 6 b

4 844 Hepatol Int (2011) 5: for Statistical Computing, Vienna, Austria) were used for analyses. All reported P values are two-sided, and are not corrected for multiple testing. P \ 0.05 was considered statistically significant. Results Apart from the serum sodium concentrations, donor and recipient baseline characteristics were similar between patients with a pre-operative serum sodium meq/ dl and all patients (Table 1). Kaplan Meier analysis showed a similar pattern of events over time for graft failure and all-cause mortality (Fig. 1, Table 2). When assuming a linear relationship with outcome, the MELD and sodium MELD scores were both significant predictors in univariable Cox regression models. For each outcome, the MELD score was slightly (descriptively) better than the sodium MELD score in terms of predictive ability and association. However, all the CPE statistics were only marginally greater than 0.5 (association by chance) and all the HR were small (estimations range from 1.05 to 1.14 per 10 units increase in the predictor), indicating poor prediction and weak association (Table 3). The P-spline plots (Figs. 2, 3) suggested that the HR for the sodium MELD score on both graft failure and all-cause mortality had a different pattern below and above 23, whereas the HR for the MELD score changed in pattern at approximately 16 and 35. Based on the maximum likelihood algorithm, the suggested cut point was 19 for the sodium MELD score, and 18 and 30 for the MELD score. We incorporated these latter cut points in piecewise models. Risk of each outcome significantly decreased for sodium MELD score increases between 6.43 (the observed lowest sodium MELD score) and 19, but the risk of each outcome increased as the score increased beyond 19. Likewise, risk decreased for increase in MELD score between 6.43 (the observed lowest MELD score) and 18, but increased as MELD score increased between 18 and 35 for either outcome (Table 4). However, the predictive ability of none of the piecewise models improved over the primary models reported above. In addition, there was no difference in graft failure (P = 0.91) or all-cause mortality (P = 0.85) across the two sodium MELD categories. However, the risk of each outcome increased across the three MELD score categories (from \18, 18 30, to [30) (P \ for each outcome, Table 5). Furthermore, the abilities of MELD score and the sodium MELD score were descriptively better (but not statistically significant) in predicting a 30-day graft failure or mortality event than long-term outcomes (Table 6). Fig. 1 a d Plot of the Kaplan Meier density function estimates and their associated equal-precision 95% confidence. a b Graft failure and all-cause mortality Kaplan Meier curves for patients with serum sodium concentration between 120 and 135 (meq/dl); c d Graft failure and all-cause mortality Kaplan Meier curves for all patients

5 Hepatol Int (2011) 5: Table 2 Summary of the number of events at various post-transplant time points Patients with Serum Sodium meq/dl (N = 6,186 a ) All patients (N = 15,156 b ) Graft failure Mortality Graft failure Mortality 1 month 413/0/ /0/ /0/ /0/ months 535/0/ /0/ /0/ /0/ months 641/0/ /0/ /0/ /0/ months 870/0/ /0/ /0/ /0/ year 1126/1439/ /1452/ /3618/ /3645/ years 1363/3250/ /3263/ /8023/ /8063/ years 1432/4440/292 0/4456/ /10899/ /10939/737 Summaries are presented as # events/# censored/# at risk a 22 and 202 patients had missing the graft failure outcome and the mortality outcome, respectively b 60 and 472 patients had missing the graft failure outcome and the mortality outcome, respectively Table 3 Primary results: association between graft failure/all cause mortality (censored at 3 years) and MELD/sodium MELD score assuming linear relationship with outcomes Predictors Graft failure All-cause mortality Hazard ratio a (95% CI) P 2 CPE b (95% CI) Hazard ratio a (95% CI) P 2 CPE b (95% CI) Patients with serum sodium (120, 135) meq/dl (N = 6,186) d Sodium MELD score c 1.05 (1.00, 1.10) (0.50, 0.53) 1.07 (1.02, 1.12) (0.50, 0.54) MELD score 1.10 (1.04, 1.17) (0.51, 0.54) 1.14 (1.07, 1.21) \ (0.52, 0.55) All patients (N = 15,156) e MELD score 1.17 (1.13, 1.21) \ (0.53, 0.55) 1.21 (1.17, 1.26) \ (0.54, 0.56) a Increased risk of graft failure/all-cause mortality for every 10 units increase in the predictor b CPE = concordance probability estimate ranges from 0.50 (chance) to 1.0 (perfect prediction) c Sodium MELD score = MELD? 1.59 (135 serum sodium) [2] d 22 and 202 patients had missing the graft failure outcome and the mortality outcome, respectively e 60 and 472 patients had missing the graft failure outcome and the mortality outcome, respectively Fig. 2 Plot of Log hazard ratio for graft failure and all-cause mortality within postoperative 3 years versus the P-spline smoothed recipient sodium MELD score for patients with serum sodium concentration between 120 and 135 (meq/dl). Dotted lines represent the 95% CI No interaction between either of the predictors (i.e., MELD and sodium MELD scores) and any of the recipient and donor variables were found to be significant for either of the outcomes at the 0.05 level. Discussion Identifying factors associated with postoperative morbidity and mortality after OLT remain of paramount importance,

6 846 Hepatol Int (2011) 5: Fig. 3 Plot of Log hazard ratio for graft failure and all-cause mortality within postoperative 3 years versus the P-spline smoothed recipient MELD score for all patients. Dotted lines represent the 95% CI Table 4 Secondary results: association between graft failure/all cause mortality (censored at 3 years) and MELD/sodium MELD from piecewise models Graft failure All-cause mortality Predictors Hazard ratio a (95% CI) P CPE b (95% CI) Hazard ratio a (95% CI) P CPE b (95% CI) Sodium MELD score c patients with serum sodium (120,135) meq/dl (N = 6,186) d \ (0.58, 0.95) (0.51, 0.54) 0.73 (0.56, 0.95) (0.51, 0.54) C (1.04, 1.17) (1.06, 1.19) \0.001 MELD score all patients (N = 15,156) e \ (0.76, 0.97) (0.54, 0.55) 0.81 (0.71, 0.92) (0.55, 0.56) (18, 30) 1.41 (1.27, 1.55) \ (1.40, 1.74) \0.001 [ (1.01, 1.26) (1.01, 1.28) 0.03 a Increased risk of graft failure/all-cause mortality for every 10 units increase in the predictor b CPE = Concordance probability estimate ranges from 0.50 (chance) to 1.0 (perfect prediction) c Sodium MELD score = MELD? 1.59 (135 - serum sodium) [2] d 22 and 202 patients had missing the graft failure outcome and the mortality outcome, respectively e 60 and 472 patients had missing the graft failure outcome and the mortality outcome, respectively Table 5 Secondary results: association between MELD score categories and graft failure/all cause mortality (censored at 3 years) MELD score Graft failure (N = 15,096) Mortality (N = 14,684) Hazard ratio a (95% CI) P Hazard ratio a (95% CI) P \ (0.82, 0.94) (0.78, 0.91) (18, 30) 1.00 (reference) 1.00 (reference) [ (1.24, 1.51) \ (1.29, 1.60) \0.001 a Relative to the MELD score category of given the shortage of available organs and the consequent need to optimize organ s allocation and resources. While no definitive predictor of outcome has been identified, severity of disease at the time of OLT is thought to be associated with the survival [7, 19]. Unsurprisingly, the MELD score has been considered as a potential predictor of post-transplant mortality and resources utilization, however, studies investigating that revealed contradicting results [5 8, 11, 20]. An investigation by Lally et al. [11], used MELD, the CTP score, and a region 1 continuous medical severity score (CMSS) to predict post-liver transplant survival in 376 patients. These investigators did not find a significant association between any of the scores and post-transplant

7 Hepatol Int (2011) 5: Table 6 Secondary results: association between the 30-day graft failure/all cause mortality and MELD/sodium MELD score Graft failure All-cause mortality Predictors Hazards ratio a (95% CI) P CPE b (95% CI) Hazards ratio a (95% CI) P CPE b (95% CI) Patients with serum sodium (120, 135) meq/dl (N = 6,186) d Sodium MELD score c 1.13 (1.04, 1.23) (0.51, 0.56) 1.26 (1.14, 1.40) \ (0.54, 0.60) MELD 1.22 (1.09, 1.35) \ (0.52, 0.57) 1.43 (1.26, 1.63) \ (0.56, 0.61) All patients (N = 15,156) e MELD 1.26 (1.19, 1.34) \ (0.54, 0.57) 1.44 (1.34, 1.54) \ (0.58, 0.61) a Increased risk of graft failure/all-cause mortality for every 10 units increase in the predictor b CPE = concordance probability estimate ranges from 0.50 (chance) to 1.0 (perfect prediction) c Sodium MELD score = MELD? 1.59 (135-serum sodium) [2] d 22 and 202 patients had missing the graft failure outcome and the mortality outcome, respectively e 60 and 472 patients had missing the graft failure outcome and the mortality outcome, respectively survival. Brown et al. [6] examined 42 consecutive patients (of 2A UNOS status) undergoing liver transplantation for non-fulminant liver disease; again, neither MELD nor CTP scores predicted post-transplant survival. Interestingly, the results of that study did not identify a significant association between surgical times, intraoperative transfusion requirement or the MELD score [6]. Few other studies demonstrated association of pre-transplant MELD with short- and long-term survival after liver transplant in particular when comparing patients with high pre-transplant MELD ([25) and low MELD score (\15) [8]. Habib et al. [20] first evaluated the long-term post-liver transplant survival in 1,472 patients using MELD. A pre-olt MELD [26 was associated with poorer short- and long-term (10 years) post-liver transplant survival. Contrary to the previous studies, our investigation included a large number of patients with reliably collected and defined preoperative variables and outcomes (UNOS database). Our analysis demonstrated that the MELD score was a significant predictor of post-transplant mortality and graft failure, but the associations were weak; furthermore, the predictive abilities of both outcomes were very poor. In addition, our study showed that the higher pre-transplant MELD score was associated with poorer post-liver transplant outcomes when comparing patients with lower pre-transplant MELD (\18, 18 30, and [30). Despite the previous reports, our findings suggest that from the practical standpoint the MELD score cannot be used to forecast post-transplant outcomes. Interestingly, MELD score may have better sensitivity to predict post-transplant outcomes in patients with hepatitis C-induced end-stage liver disease. Onaca et al. [12] found that pre-transplant MELD score above 24 in patients with hepatitis C-induced liver disease was associated with significantly lower survival, questioning transplant suitability for these patients. Wang et al. [10] found in a small study that MELD score was an objective predictive system and more efficient than CTP score to evaluate the risk of 3-month morbidity and short-term prognosis in patients with acute-on-chronic hepatitis B liver failure undergoing OLT. On the contrary, our findings did not show any interaction between either of the two MELD scores and either cause of the ESLD or other recipient and donor variables. Our results for these interaction effects, although not our primary aim, are more reliable than other reports based on relatively smaller sample sizes. Furthermore, it is difficult to compare among studies as the use of recipient and donor variables is inconsistent, and so are the reported outcomes. There is a strong data suggesting the importance of the serum sodium concentration on pre-transplant outcomes: Ruf et al. [4] made observation by incorporating serum sodium concentration into the MELD score, renal impairment and/or circulatory dysfunction that can be identified at an earlier phase in patients with advanced cirrhosis. Investigations by Biggins et al. and Kim et al. [3, 4], respectively, further confirmed that addition of serum sodium into MELD score offers more sensitive tool than MELD score alone to predict 90 days waiting list mortality. Kim et al. [3] concluded that potentially as many as 7% of waiting-list deaths could be prevented, if the sodium MELD score had been utilized to allocate the organs. Although hyponatremia at the time of liver transplantation has been associated with increased post-transplant morbidity and mortality in some studies, it remains unclear whether this association is attributable to hyponatremia itself, or reflects the fact that patients with hyponatremia tend to be sicker than their normonatremic counterparts [13, 14, 21, 22]. But it is plausible to assume that incorporation on pre-transplant serum sodium with MELD score may provide a better tool to predict post-transplant outcomes. Only one large study by Dawwas et al. [14] assessed the effect of recipient sodium on the survival of adults with

8 848 Hepatol Int (2011) 5: chronic liver disease in a cohort who received a liver transplant (UK and Ireland during the period from March 1, 1994 to March 31, 2005 [n = 5,152]) while adjusting for a wide range of recipient, donor, and graft characteristics. The investigators found that severe recipient hyponatremia (serum sodium \130 meq/dl) was associated with an increased 90-day post-transplant mortality [14]. In our investigation, despite statistically significant association between the sodium MELD score and the post-transplant outcomes in patients with pre-transplant serum sodium between 120 and 135 meq/dl, the predictive abilities were marginal. To our knowledge, this investigation is the first attempt to evaluate and compare predictive ability of sodium MELD versus traditional MELD score as predictors of post-liver transplant outcomes. We found that the abilities of both MELD and sodium MELD scores to predict survival and graft failure after OLT surgery were low. Weak CPE values suggest that outcomes after OLT are largely influenced by surgical factors and immediate postoperative management rather than preoperative recipient and donor status (represented by the MELD score, recipient serum sodium, DRI, etc.). An obvious limitation of our study is its retrospective nature and the fact that we had to exclude a substantial number of patients who had incomplete data. Some potentially important factors affecting postoperative outcome were not included in the analyses because they were not available in the UNOS data set in particular, intraoperative and immediate postoperative variables. However, the major danger of this sort of retrospective analyses is over-estimating the predictive ability. Our major conclusion is that although MELD and sodium MELD were significantly associated with outcome, the predictive abilities were too low to be clinically useful; it seems quite unlikely that a prospective study would provide additional value. Our results are an example of a finding that statistically significant values due to a large sample size are not clinically useful. In conclusion, we found that the sodium MELD and the traditional MELD scores were associated with postoperative graft failure and mortality, but that their predictive abilities of these outcomes were poor. The modest CPE values of these two scoring systems may result from the fact that other important intraoperative and postoperative factors have substantial influence on post-transplant graft failure and mortality. Acknowledgements This work was supported in part by Health Resources and Services Administration contract C. The content is the responsibility of the authors alone and does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the US Government. References 1. Freeman RB, Wiesner RH, Edwards E, Harper A, Merion R, Wolfe R, United Network for Organ Sharing Organ Procurement and Transplantation Newtwork Liver and Transplatation Committee. Results of the first year of the new liver allocation plan. Liver Transpl 2004;10: Biggins SW, Kim WR, Terrault NA, Saab S, Balan V, Schiano T, Benson J, Therneau T, Kremers W, Wiesner R, Kamath P, Klintmalm G. Evidence-based incorporation of serum sodium concentration into MELD. Gastroenterology 2006;130: Kim WR, Biggins SW, Kremers WK, Wiesner RH, Kamath PS, Benson JT, Edwards E, Therneau TM. Hyponatremia and mortality among patients on the liver-transplant waiting list. N Engl J Med 2008;359: Ruf AE, Kremers WK, Chavez LL, Descalzi VI, Podesta LG, Villamil FG. Addition of serum sodium into the MELD score predicts waiting list mortality better than MELD alone. Liver Transpl 2005;11: Kim DJ, Lee SK, Jo JW, Kim SJ, Kwon CHD, Park JW, Han YS, Park JB. Prognosis after liver transplantation predicted by preoperative MELD score. Transplant Proc 2006;38: Brown RS Jr, Kumar KS, Russo MW, Kinkhabwala M, Rudow DL, Harren P, Lobritto S, Emond JC. Model for end-stage liver disease and Child-Turcotte-Pugh score as predictors of pretransplantation disease severity, posttransplantation outcome, and resource utilization in United Network for Organ Sharing status 2A patients. Liver Transpl 2002;8: Habib S, Berk B, Chang CC, Demetris AJ, Fontes P, Dvorchik I, Eghtesad B, Marcos A, Shakil AO. MELD and prediction of postliver transplantation survival. Liver Transpl 2006;12: Onaca NN, Levy MF, Sanchez EQ, Chinnakotla S, Fasola CG, Thomas MJ, Weinstein JS, Murray NG, Goldstein RM, Klintmalm GB. A correlation between the pretransplantation MELD score and mortality in the first two years after liver transplantation. Liver Transpl 2003;9: Brandao A, Fuchs SC, Gleisner AL, Marroni C, Zanotelli ML, Cantisani G, Liver Transplantation Group. MELD and other predictors of survival after liver transplantation. Clin Transplant 2009;23: Wang ZX, Yan LN, Wang WT, Xu MQ, Yang JY. Impact of pretransplant MELD score on posttransplant outcome in orthotopic liver transplantation for patients with acute-on-chronic hepatitis B liver failure. Transplant Proc 2007;39: Lally A, Nixon A, Lewis D, Pomfret E, Pomposelli J, Fabry S, Goldberg E, Jenkins R, Gordon F. MELD, CTP, and region 1 CMSS equally predict post liver transplant patient survival (abstract). Hepatology 2001;34:290A 12. Onaca NN, Levy MF, Netto GJ, Thomas MJ, Sanchez EQ, Chinnakotla S, Fasola CG, Weinstein JS, Murray N, Goldstein RM, Klintmalm GB. Pretransplant MELD score as a predictor of outcome after liver transplantation for chronic hepatitis C. Am J Transplant 2003;3: Londono M, Guevara M, Rimola A, Navasa M, Taura P, Mas A, Garcia-Valdecasas J, Arroyo V, Gines P. Hyponatremia impairs early posttransplantation outcome in patients with cirrhosis undergoing liver transplantation. Gastroenterology 2006;130: Dawwas MF, Lewsey JD, Neuberger JM, Gimson AE. The impact of serum sodium concentration on mortality after liver transplantation: a cohort multicenter study. Liver Transpl 2007;13: United Network for Organ Sharing Organ data STAR file from August 20. Data Use Agreement-I. Available from:

9 Hepatol Int (2011) 5: agreementpdf 2008; This work was supported in part by Health Resources and Services Administration contract C. The content is the responsibility of the authors alone and does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the US Government 16. Cox D. Regression models and life tables. J R Stat Soc Ser B Stat Methodol 1972;34: Gonen M, Heller G. Concordance probability and discriminatory power in proportional hazards regression. Biometrika 2005;92: Kaplan E, Meier P. Nonparametric estimation from incomplete observations. J Am Stat Assoc 1958;53: Cholongitas E, Marelli L, Shusang V, Senzolo M, Rolles K, Patch D, Burroughs AK. A systematic review of the performance of the model for end-stage liver disease (MELD) in the setting of liver transplantation. Liver Transpl 2006;12: Habib S, Dvorchik I, Ahmad J, Chopra K, Fontes P, Marcos A, Fung J, Shakil AO. MELD as predictor of post-transplantation survival. (abstract). Hepatology 2004;40(Suppl 1):261A 21. Wijdicks EF, Blue PR, Steers JL, Wiesner RH. Central pontine myelinolysis with stupor alone after orthotopic liver transplantation. Liver Transpl Surg 1996;2: Bronster DJ, Emre S, Boccagni P, Sheiner PA, Schwartz ME, Miller CM. Central nervous system complications in liver transplant recipients incidence, timing, and long-term followup. Clin Transplant 2000;14:1 7