Effects of Allocating Livers for Transplantation Based on Model for End-stage Liver Disease-Sodium Scores on Patient Outcomes

Transcription

1 Accepted Manuscript Effects of Allocating Livers for Transplantation Based on Model for End-stage Liver Disease-Sodium Scores on Patient Outcomes Shunji Nagai, MD, PhD, Lucy C Chau, HBSc, MMI, Randolph E Schilke, MPH, Mohamed Safwan, MBBS, Michael Rizzari, MD, Kelly Collins, MD, Atsushi Yoshida, MD, FACS, Marwan S Abouljoud, MD, FACS, Dilip Moonka, MD PII: S (18) DOI: /jgastro Reference: YGAST To appear in: Gastroenterology Accepted Date: 20 July 2018 Please cite this article as: Nagai S, Chau LC, Schilke RE, Safwan M, Rizzari M, Collins K, Yoshida A, Abouljoud MS, Moonka D, Effects of Allocating Livers for Transplantation Based on Model for Endstage Liver Disease-Sodium Scores on Patient Outcomes, Gastroenterology (2018), doi: / jgastro This is a PDF file of an unedited manuscript that has been accepted for publication As a service to our customers we are providing this early version of the manuscript The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain

2 Waitlist mortality and transplant in the MELD-Na and MELD periods (Propensity score matching model) Transplant rate Mortality rate MELD-Na period MELD period MELD period MELD-Na period Hazard Ratio Survival benefit of liver transplant in the MELD-Na and MELD periods Score category MELD-Na Era period (WL+TS) MELD Era period (WL+TS)

3 1 Title: Effects of Allocating Livers for Transplantation Based on Model for End-stage Liver Disease-Sodium Scores on Patient Outcomes (Short title: MELD-Na Based Allocation in Liver Transplant) Shunji Nagai, MD, PhD* 1 Lucy C Chau, HBSc, MMI* 1 Randolph E Schilke, MPH 1 Mohamed Safwan, MBBS 1 Michael Rizzari, MD 1 Kelly Collins, MD 1 Atsushi Yoshida, MD, FACS 1 Marwan S Abouljoud, MD, FACS 1 Dilip Moonka, MD 2 * Share first authorship 1 Transplant and Hepatobiliary Surgery, Henry Ford Hospital, Detroit, Michigan, United States 2 Gastroenterology, Henry Ford Hospital, Detroit, Michigan, United States Abbreviations: model for End-Stage Liver Disease-Sodium (MELD-Na); liver transplant (LT); Organ Procurement and Transplantation Network (OPTN); the Standard Transplant Analysis and Research (STAR); hepatocellular carcinoma (HCC); hepatitis C virus (HCV); hazard ratio (HR); confidence interval (CI)

4 2 Correspondence and requests for reprints Shunji Nagai, MD, PhD Transplant and Hepatobiliary Surgery, Henry Ford Hospital, 2799 W Grand Blvd CFP-2 Detroit, MI Phone: Fax: snagai1@hfhsorg Disclosures: Nothing to disclose This work received no funding from any agency in the public, commercial, or not-for-profit sectors No writing assistance Author contributions Study concept and design: Nagai, Chau, Moonka, Safwan, Yoshida, Abouljoud Acquisition and interpretation of data: Nagai, Chau, Safwan, Schilke, Collins, Rizzari Drafting of the manuscript: Nagai, Chau, Safwan, Schilke, Moonka Critical revision of the manuscript: Collins, Rizzari, Yoshida, Abouljoud Statistical analysis: Nagai, Chau, Schilke Study supervision: Moonka, Yoshida, Abouljoud

5 3 ABSTRACT Background & Aims: The model for end-stage liver disease-sodium (MELD-Na) score was introduced for liver allocation in January 2016 We evaluated the effects of liver allocation, based on MELD-Na score, on waitlist and post-transplant outcomes Methods: We examined 2 patient groups from the United Network for Organ Sharing registry; the MELD-period group comprised patients who were registered as transplant candidates from June 18, 2013 through January 10, 2016 (n=18,850) and the MELD-Na period group comprised patients who were registered from January 11, 2016 through September 30, 2017 (n=14,512) We compared waitlist and post-transplant outcomes and association with serum sodium concentrations between groups Results: Mortality within 90 days on the liver waitlist decreased (hazard ratio [HR], 0738; P<001) and transplant probability increased significantly in the MELD-Na period (HR, 1217; P<001) Although mild, moderate, and severe hyponatremia ( mmol/l, mmol/l, and <125 mmol/l) were independent risk factors for waitlist mortality in the MELD period (HR, 1354, 1762, and 2656; P<001, P<001, and P<001, respectively), compared to the reference standard ( mmol/l), these adverse outcomes were reduced in the MELD- Na period (HR, 1092, 1271 and 1374; P=27, P=018, and P=037, respectively) The adjusted survival benefit of transplant vs individuals placed on the waitlist in the same score categories was definitive for patients with MELD-Na scores of in the MELD-Na era (HR, 0336; P<001) compared to MELD scores of in the MELD era (HR, 0365; P<001)

6 4 Conclusions: The MELD-Na score based liver allocation successfully improved waitlist outcomes and provided significant benefit to hyponatremic patients Given the discrepancy in transplant survival benefit, the current rules for liver allocation might require revision Keywords: Hyponatremia; Waitlist mortality; Survival benefit; United Network for Organ Sharing

7 5 TEXT The model for End-Stage Liver Disease (MELD) score had been used for the allocation of deceased donor liver grafts in the United States since , 2 The MELD score predicts mortality in patients with liver cirrhosis and prioritizes patients for liver transplant (LT) on that basis However, hyponatremia is an independent prognostic factor in patients with cirrhosis but was not included in the original MELD score calculation 3-5 Kim et al proposed a modified scoring system to further reduce waitlist mortality that incorporates serum sodium concentration into the MELD equation 6 This MELD-Na score was officially implemented for liver graft allocation in January In the United States, the liver allocation system includes Share 15 and Share 35 rules 2 The Share 35 rule offers livers to both local and regional patients on the wait list with MELD (MELD-Na) scores of 35 or higher The Share 15 rule offers liver grafts from deceased donors regionally or nationally before returning to local candidates with MELD (MELD-Na) <15 Despite implementation of the new allocation using the MELD-Na score, the regional share systems based on specific scores (Share 15 and Share 35) have not been reevaluated These rules were determined using data from the MELD period 7, 8 The rationale of Share 15 was derived from the results of a study conducted by Merion et al, 9 that evaluated the survival benefit of LT based on MELD score However, the impact on MELD-Na based allocation on waitlist and posttransplant outcomes has not previously been investigated To better understand the effects of the new liver allocation based on MELD-Na score, we compared waitlist outcomes and post-transplant outcomes between the MELD and MELD-Na periods after Share 35 was implemented In addition, we reevaluated the survival benefit of liver transplantation in the MELD-Na period It should be noted that the two major studies by Kim et

8 6 al (the original study for MELD-Na score proposal) and Merion et al (the original study for survival benefit of liver transplantation) used different endpoints for waitlist outcome 6, 9 The study by Kim et al proposed MELD-Na score by assessing 90-day waitlist outcome The study by Merion et al investigated survival benefit of liver transplant by comparing overall waitlist mortality and one-year post-transplant patient survival In this study, we mainly investigated 90- day waitlist outcomes to assess the impact of the MELD-Na based allocation, whereas overall waitlist mortality was evaluated to compare survival benefit of liver transplant in the MELD and MELD-Na periods

9 7 Methods Patient selection This study used data from the Organ Procurement and Transplantation Network (OPTN) / United Network for Organ Sharing (UNOS) contained in the Standard Transplant Analysis and Research (STAR) file, which included data for all patients on the liver waitlist registered up to September 30, 2017 All adult patients (18 years or older) who were registered after implementation of Share 35 (June 18, 2013) for liver or liver-kidney transplantation were eligible for inclusion in this study Exclusion criteria included age less than 18 years, retransplantation, Status 1 listing, and liver transplant combined with thoracic organ(s), pancreas, and/or intestine When patients were removed from the waitlist because a transplant was done at a different center, their waitlist data were excluded from the analysis Patients who received exception score(s) were censored on the day of exception score approval The OPTN policy for exception score(s) for hepatocellular carcinoma (HCC) was revised during the study period (October 2015) 10 ; therefore, patients with exception score(s) for HCC were excluded Patients who received living donor liver transplant were censored on the day of transplant We examined two patient groups; MELD period group composed of patients registered on the list between June 18, 2013 and January 10, 2016, and MELD-Na period group composed of patients registered between January 11, 2016 and September 30, 2017 Analysis for waitlist and post-transplant outcomes Waitlist patients were classified based on their initial laboratory MELD/MELD-Na score to analyze waitlist outcomes Transplant recipients were classified based on their final pretransplant score to analyze post-transplant outcomes These classifications were performed

10 8 according to MELD score for patients in the MELD period and MELD-Na score for those in the MELD-Na period Of note, MELD-Na score is applied when MELD score is 12 or higher 11 Patients were stratified according to MELD/MELD-Na score (12 groups) and initial serum sodium concentration range groups (5 groups; mild [ mmol/l], moderate [ mmol/l], and severe [<125 mmol/l] hyponatremia, hypernatremia [>145 mmol/l], normal sodium concentration group [ mmol/l]) The prognostic impact of hyponatremia on waitlist outcomes (90-day mortality) was evaluated in the two periods As differences in follow-up time may cause a withdrawal bias, patients registered in the MELD and MELD-Na periods were censored on the last day of each period (January 10, 2016 and September 30, 2017, respectively) in both waitlist and post-transplant outcome analyses MELD and MELD-Na score models were developed to predict waitlist mortality at 90 days after registration 6 Therefore, we used 90-day waitlist outcomes as a main end point (mortality and transplant probability) In addition, we used overall waitlist outcomes as an additional end point Regarding post-transplant outcome, one year-graft survival was set as an end point A subgroup analysis for waitlist and post-transplant outcomes was conducted according to hepatitis C virus infection (HCV) status to consider possible cofounding effects of the introduction of new directacting antivirals on waitlist and post-transplant outcomes The survival benefit of liver transplantation was compared between the two periods as a function of MELD/MELD-Na score The survival benefit is expressed as a ratio comparing the mortality rate of LT recipients with candidates on the waitlist (death per 1000 patient year) 9 The waitlisted patients were classified by their final score In this analysis, waitlist outcome throughout the period and post-transplant outcome with one-year follow-up were set as end points

11 9 Definitions of waitlist outcomes In the analysis of waitlist mortality, removal from the waitlist because of deterioration in medical condition (ie too sick to transplant) was considered as waitlist mortality Many of these patients were critically ill and die within a short period of waitlist removal On the other hand, the liver transplant survival benefit analysis was performed twice using two definitions of waitlist mortality: 1) too sick to transplant was not considered as waitlist mortality, which mirrors the methodology in the study by Merion et al 9 and 2) too sick to transplant was considered as waitlist mortality Statistical Analysis Data was summarized using the median with interquartile range (IQR) for continuous variables and using percentage for discrete variables Continuous variables were analyzed using the Mann Whitney-U test and discrete variables were analyzed using a chi-square test Transplant probability and waitlist mortality rates were compared by the Gray test to estimate cumulative incidence of competing events, including death, transplant, and other reasons for removal from the waitlist Multivariable models were created to elucidate factors that impact waitlist and posttransplant outcomes Waitlist mortality and transplant probability were compared between the MELD and MELD-Na periods using Fine-Gray proportional hazard regression models The Fine-Gray models allow for the analysis of competing risk events which in our study include mortality, transplant, and other reasons for waitlist removal Post-transplant outcomes were compared using Cox proportional hazard models The risk of waitlist mortality was adjusted for recipient age, race, gender, primary diagnosis, listed for liver-kidney transplant, UNOS region

12 10 (1-11), history of diabetes, Karnofsky score at registration lower than 30%, presence of ascites, hepatic encephalopathy, and dialysis requirement at registration The risk of liver graft loss was adjusted for recipient age, race, gender, primary diagnosis, liver-kidney transplant, UNOS region (1-11), organ share type (local, regional, or national), history of diabetes, Karnofsky score at transplantation lower than 30%, donor age, cold ischemia time, donation after cardiac death, presence of ascites, hepatic encephalopathy and dialysis requirement at transplant 8 To further validate our analysis of waitlist outcomes between the MELD and MELD-Na, we used a propensity score matching model, in addition to the multivariate analysis We estimated the propensity score by period for each waitlisted patient using a multivariable logistic regression model, in which all variables listed above were included We then used the propensity score, the single composite variable, to match patients in each period by matching the propensity score using a caliper of 020 standard deviations 14,244 patients were compared in the MELD and MELD-Na periods By using the matched groups, 90-day waitlist mortality rate and transplant probability were estimated and compared by the Gray test and hazard ratios were estimated by a Fine-Gray regression model In the survival benefit analysis, delta MELD (MELD-Na) score was computed by dividing the change in score by the length of the time interval between reported dates of change 12 Delta MELD (MELD-Na) score was included in the covariates for the multivariable model The STAR waitlist database was queried for new final MELD (MELD-Na) scores within 90 days prior to recipients censoring point, whether that be transplant, removal from waiting list, or end-of-period From the identified new final score date, we further identified prior MELD (MELD-Na) scores 25 days preceding the newly contrived new final scores We estimated the 30-day average Delta MELD as follows

13 11 = (( ) ( )) " " &" " All statistical analyses were completed using SPSS version 19 (IBM, Chicago, USA) and R (R Foundation for Statistical Computing, Vienna Austria), and significance was set at 005

14 12 Results Recipient characteristics Before applying the inclusion and exclusion criteria, the number of waitlisted and transplanted patients in the MELD and MELD-Na periods were compared The total number of patients who were added to the waitlist was 30,723 and 22,269, in the MELD period (909 days) and MELD- Na period (629 days), respectively The newly listed patients significantly increased in the MELD-Na period (1,014 to 1,062 patients/month [P <0001]) The total number of liver transplant was 17,584 and 13,759 in the MELD and MELD-Na periods The monthly number of transplant significantly increased in the MELD-Na period (580 to 656 transplants/month [P <0001]) There were 18,850 patients in the MELD period and 14,512 patients in the MELD-Na period that met inclusion criteria for the waitlist outcome analysis Of these, 6,547 and 5,862 patients who underwent primary deceased donor LT in the same period were evaluated in the post-transplant outcome analysis The actual score at registration and transplant significantly increased in the MELD-Na period, whereas the proportion of patients with poor performance status at registration (Karnofsky score <30%) and those requiring dialysis or life support decreased (Table 1) In the MELD-Na period, the score distributions shifted from lower score to mid score categories both at registration and at transplant At registration, 383% of patients had a score between in the MELD-Na period compared to 296% in the MELD period (Figure 1a) At transplant, 528% of patients had a score between in the MELD-Na period compared to 468% in the MELD period (Figure 1b) The serum sodium concentration at transplant in the MELD-Na period was significantly lower in patients with scores between 6-11 and (Figure 1c)

15 13 Comparisons of waitlist and post-transplant outcomes between the MELD and MELD-Na periods Waitlist outcomes at 90 days after registration (mortality and transplant) and one year-liver graft loss were compared between the MELD and MELD-Na periods (Figure 2a, 2b, and 2c) The hazard of 90-day waitlist mortality in the MELD-Na period was 262% lower than in the MELD period (HR =0738, 95% CI = , P <0001) The 90-day transplant probability increased by 217% in the MELD-Na period (HR =1217, 95% CI = , P <0001) The risk of one year-graft loss or patient death was similar between the MELD and MELD-Na periods (Graft loss: HR [95% CI] =1086 [ ], P =021; Patient death: HR [95% CI] =1068 [ ], P =0323) Overall waitlist outcomes were also compared, which demonstrated similar trends The overall waitlist mortality in the MELD-Na period was 335% lower than in the MELD period (HR =0665, 95% CI = , P <0001) The overall transplant probability was 197% higher in the MELD-Na period than in the MELD period (HR =1197, 95% CI = , P <0001) (sfigure 1) The hazard of 90-day waitlist mortality in the MELD-Na period significantly decreased in patients with normal serum sodium concentration (HR [95% CI] =0825 [ ], P <0001) and in those with mild (HR [95% CI] =0656 [ ], P <0001), moderate (HR [95% CI] =0640 [ ], P <0001) or severe hyponatremia (HR [95% CI] =0526, [ ], P <0001) We did propensity score matching for waitlisted patients between the MELD and MELD- Na period groups to validate the findings in the waitlist outcome analysis After the propensity score matching, 14,244 waitlisted patients were selected from each group Compared to the

16 14 MELD period group, the MELD-Na period group had significantly better 90-day waitlist outcomes with less mortality (81% vs 105%, P <0001) and greater transplant probability (348% vs 300%, P <0001) (Figure 3a) When including propensity score in a risk adjustment model, the hazards of 90-day waitlist mortality and transplant probability were 0750 [ ] (P <0001) and 1186 [ ] (P <0001), respectively The model achieved standardized difference less than 010 (10%) after the matching for all covariates (Figure 3b) Association between serum sodium concentration and waitlist mortality In the MELD period, the hazard of 90-day waitlist mortality in patients with mild, moderate and severe hyponatremia was 1354, 1762 and 2656 times those with normal sodium levels (HR [95% CI] = 1354 [ ], 1762 [ ], and 2656 [ ], P <0001, <0001 and <0001, respectively) (Figures 4a, 4b, and 4c) This adverse impact decreased significantly in the MELD-Na period (HR [95% CI] =1092 [ ], 1271 [ ] and 1374 [ ], P =027, 0018 and =0037, in those with mild, moderate and severe hyponatremia, respectively) (Figures 4d, 4e, and 4f) The risk of waitlist mortality associated with hyponatremia was more prominent in patients in the low-middle score categories (score of 6-26) in the MELD period, however, this was no longer observed in the MELD-Na period Waitlist and post-transplant outcomes according to Hepatitis C virus infection status In the non-hcv group, significant improvement of waitlist outcomes was observed in the MELD-Na period The hazard of 90-day waitlist mortality was 282% lower in the MELD-Na period (HR [95% CI] =0718 [ ], P <0001), and the likelihood of transplant increased by 208% (HR [95% CI] =1208 [ ], P <0001) In the HCV group, the risk of 90-day

17 15 waitlist mortality was comparable between the two groups (HR [95% CI] =0854 [ ], P =013), but the 90-day transplant probability increased by 263% in the MELD-Na period (HR [95% CI] =1263 [ ], P <0001) Comparing the MELD and MELD-Na period, there was no difference in one year-graft loss for both the non-hcv patients (HR [95% CI] =1092 [ ], P =0224) and the HCV patients (HR [95% CI] =1018 [ ], P =0913) Patient characteristics between the HCV and non-hcv groups were compared In the MELD period, the score distribution at registration in the lower (6-15), middle (16-34), and higher score (35+) categories was 330%, 569%, 101% in the HCV group and 288%, 596%, 116% in the non-hcv group, respectively In the MELD-Na period, the distribution in these three categories were 293%, 622%, and 85% in the HCV group and 230%, 654%, and 116% in the non-hcv group, respectively The odds for patients with HCV being listed in the lower score category compared with those with non-hcv, increased from 122 (95% CI = , P <0001) to 1388 (95% CI = , P <0001), whereas the odds for being listed in the higher score category decreased from 0852 (95% CI = , P =0005) to 0709 (95% CI = , P <0001) Survival benefit of liver transplantation in the MELD and MELD-Na periods The overall waitlist mortality rates were 1692 and 1707 deaths per 1000 patient years in the MELD-Na and MELD period, respectively The mortality rates among transplanted individuals with one-year follow-up were 1141 and 983 deaths per 1000 patient years in the MELD-Na and MELD period, respectively (Table 2) Figure 5 demonstrates the potential benefit of transplant described in Merion et al 9 by determining an adjusted mortality risk after transplant per 1000 patient years vs the waitlist mortality per 1000 patient years A hazard ratio less than 10 favors

18 16 transplant A transition point of liver transplant survival benefit shifted towards a higher score category in the MELD-Na period The adjusted survival benefit of transplant over individuals waitlisted in the same score categories was definitive in those with MELD-Na 24 in the MELD- Na period (score 24-26: HR [95% CI] =0563 [ ], P =0021) compared to MELD 18 in the MELD period (score 18-20: HR 95% CI =0484 [ ], P =0007) (Figure 5a) When considering removal for too sick to transplant as waitlist mortality, transplant became more favorable but the same trend was observed in a comparison of survival benefit between the periods The rates for waitlist mortality including removal for too sick to transplant were 3070 and 3182 per 1000 patients years in the MELD-Na and MELD periods, respectively (Table 2) Survival benefit of liver transplant was definitive in those with score category of in the MELD period (HR [95% CI] =0365 [ ], P <0001), whereas this shifted towards the higher score category of in the MELD-Na period (HR [95% CI] =0336 [ ], P <0001) (Figure 5b) Share 15 and Share 35 in the MELD-Na period Hyponatremia drives MELD-Na scores which might benefit hyponatremic patients whose MELD score did not meet the cut off for liver regional share (15 or 35) Patients who had an initial and final MELD <15 but MELD-Na 15 had a significantly greater likelihood of transplant within 90 days in the MELD-Na period at 261% vs 127% in the MELD period (P <0001), and they had significantly lower 90-day waitlist mortality in the MELD-Na period at 19% vs 106% in the MELD period (P <0001) (sfigure 2a) The adjusted 90-day transplant probability was 1250% higher (HR =2250, 95% CI = , P <0001), and the adjusted hazard of 90-day waitlist mortality was 709% lower (HR =0291, 95% CI = , P

19 17 <0001) in the MELD-Na period Waiting time till transplant was significantly shorter in the MELD-Na period (39 days vs 73 days, P =0002) (stable 1) In patients whose initial MELD score was <35 but MELD-Na score was 35, the likelihood of transplant within 90 days was significantly higher in the MELD-Na period at 805% vs 702% in the MELD period (P =0002) (sfigure 2b) There was a decreased 90-day waitlist mortality at 143% vs 239% respectively (P =0023) The adjusted 90-day transplant probability was 673% higher in the MELD-Na period (HR =1673, 95% CI = , P <0001), and the adjusted hazard of 90-day waitlist mortality was 372% lower in the MELD-Na period, but the difference did not reach statistical significance (HR =0628, 95% CI = , P =0062) Waiting time till transplant was significantly shorter in the MELD-Na period (6 days vs 9 days, P =0012) (stable 1) In patients with both an initial MELD and MELD-Na 35, the 90-day transplant probability increased from 747% in the MELD-Na period vs 714% in the MELD period (P =0019), and 90-day waitlist morality rate decreased significantly to 222% vs 257% (P =0006) The adjusted 90-day transplant probability was 119% higher (HR =1119, 95% CI = , P =0006), and the adjusted hazard of 90-day waitlist mortality was 189% lower (HR =0811, 95% CI = , P =0005) in the MELD-Na period (sfigure 2c)

20 18 Discussion The MELD-Na score was expected to provide better calibration and discrimination of waitlist 6, 13 mortality This study demonstrated that waitlist outcomes significantly improved in the MELD-Na period, compared to the MELD period We mainly investigated 90-day waitlist outcomes because MELD-Na score was originally proposed by Kim et al using this endpoint 6 Of note, this differs from the methodology by Merion et al where survival benefit of liver transplant was evaluated using overall waitlist mortality and one year-post-transplant survival 9 When overall waitlist outcomes were assessed in our cohort, similar trends to the results of the analysis for 90-day outcomes were observed (sfigure 1) While the number of transplanted patients significantly increased in the MELD-Na period, the number of newly listed patients also significantly increased The recent increase in the donor pool may be associated with the opioid epidemic and the increased liver graft availability might contribute to the improved waitlist outcome 14, 15 If the increased donor availability in the MELD-Na was the only reason for the findings in this study, the positive impact should have been equally observed across all score categories However, there was a discrepancy of impact and the positive impact was more prominent in the lower score categories In addition, the risk of hyponatremia for waitlist mortality clearly decreased in the MELD-Na period, which cannot be explained by the increased number of liver graft These results indicate that liver graft availability was not the main reason for decreased waitlist mortality observed in this study The MELD-Na score gives between 1 and 11 additional points; and maximum additional points are given to patients with low MELD score and severe hyponatremia 6 Our results suggest that higher scores secondary to hyponatremia probably led to better waitlist outcomes especially in the population who benefited from Share 15 (ie patients with MELD <15 but MELD-Na 15)

21 19 As expected, the MELD-Na based allocation helped alleviate the adverse prognostic impact of hyponatremia on waitlist outcomes 16 Hyponatremia is a well-known factor associated with mortality in liver cirrhosis patients and this was clearly observed in the MELD period 3, 6 The risk of mild hyponatremia disappeared in the MELD-Na period Moderate and severe hyponatremia was still significant risk factors, but the hazards decreased by 279% (1762 to 1271) and 483% (2656 to 1374) in the MELD-Na period These findings support the rationale for the implementation of MELD-Na score into liver allocation The availability of effective antiviral therapy for HCV in recent years is a potential reason for the improved waitlist outcomes observed in this study 17, 18 However, the subgroup analysis of HCV and non-hcv patients revealed that the improvement in waitlist outcomes was more obvious in non-hcv patients than those with HCV While both groups had a higher likelihood of transplant in the MELD-Na period, only the non-hcv patients had improved waitlist mortality Direct antigen antivirals have been clinically available since 2013, and the MELD-Na period began in January of 2016 The improved antiviral therapy may have already helped stabilize patients with HCV and decreased their mortality on the waitlist prior to MELD- Na implementation 19 Hence, the positive impact of new allocation on the waitlist outcomes in the HCV population was not as obvious as in the non-hcv population The impact of the MELD-Na score was less pronounced in patients with higher scores In part, this is because the additional scores added for hyponatremia are smaller in higher MELD score groups While patients whose MELD-Na score was increased by hyponatremia to over 35 did benefit with the MELD-Na period, the benefit observed was less obvious than in patients in lower score categories 20, 21 These results suggest that the MELD-Na based allocation potentially provided more benefits to patients in the less medically urgent status categories Sharma et al 22

22 20 suggested that additional scores by serum sodium concentration should be considered for candidates with a MELD score of at least 12, and MELD-Na is currently applied to patients whose MELD score is 12 or higher 11 The MELD-Na score based allocation might create more competition in the mid-high score categories As such, the benefit of MELD-Na based allocation might be diminished in the higher score category Because of the tighter competition within this patient group, new criteria to discriminate medical urgency in the sicker patient population may need to be established The Share 15 policy mandates that if there is no local candidate with MELD score 15, organs are offered regionally or nationally before returning to local candidates with MELD <15 2 In this study, we found that in the MELD-Na period, the survival benefit of liver transplant diminished in the lower score categories, and a transition point justifying survival benefit shifted towards higher score categories This transition might result from the combination of the improved waitlist outcomes and worsened post-transplant outcomes in the MELD-Na period (Death per 1000 Patient Years in Table 2) The improved waitlist outcomes can be explained by the positive impact of the MELD-Na based allocation, as discussed before Reasons for the slightly worsened post-transplant outcomes in the MELD-Na period remain unclear However, we have observed that the liver transplant population has become older Increased recipient age is a known poor prognostic factor in liver transplant recipient and might be associated with the findings observed in this study 26 As such, the current rules for liver allocation may be suboptimal under the MELD-Na based allocation and the criteria for Share 15 may need to be revisited Based on our findings, the cut-off score for share might be raised to a score of about 21, as Share 21 (Hazard become less than 10 in the score category of 21-23) (Figure 5a) Of note, our methodology in the survival benefit analysis followed that of Merion et al where waitlist

23 21 mortality was defined as death on the waitlist 9 Aside from the survival benefit analysis, waitlist mortality in this study was defined as death on the waitlist or removal for reason of too sick to transplant In a subgroup analysis, the survival benefit was recalibrated to consider removal for too sick to transplant as waitlist mortality Similar trends were still observed where the survival benefit of transplant shifted to high score ranges in the MELD-Na period (Figure 5b) According to this analysis, the survival benefit of liver transplant was definitive in patients with score category of 21-23, which could further validate our proposal to revise Share 15 rule to Share 21 The findings of our study may trigger further discussions of liver graft prioritization According to the Final Rule, the priority rankings shall be ordered from most to least medically urgent 27 Liver transplantation in the United States is represented by the OPTN/UNOS liver/intestinal board which prioritizes transplantation access to sickest population Share 35 rule was designed and implemented in 2013 to fulfill this requirement by sharing liver graft within a region specifically for patients with very high MELD/MELD-Na scores (35 or higher) Recently, the OPTN/UNOS Liver/Intestine Board has approved of a new enhanced liver distribution system One of the criteria in the new liver allocation policy is the following 28 Additional transplant priority (equivalent to 3 MELD or PELD points) will be awarded to liver candidates with a MELD or PELD of at least 15, and who are either within the same donor service area as a liver donor or are within 150 nautical miles of the donor hospital but in a different donor service area The Share 15 rule, which was based on the study by Merion et al, has not been reviewed for over a decade but is still maintained in upcoming policy 9 Our study revealed that the MELD-Na based allocation provided significant benefit to hyponatremic patients with lower MELD score

24 22 and that survival benefit of liver transplant in patients with MELD-Na score between became unclear and questionable in the MELD-Na period The Final Rule stated, Allocation policies shall be reviewed periodically and revised as appropriate We propose a new Share 21 by raising the cut-off score from 15 to 21 based on our findings We believe this will eventually promote sicker patient access to transplant by allocating the liver efficiently The effect of the MELD-Na based allocation in patients with exception points remains to be elucidated In this study, patients with HCC were excluded because the policy on HCC exception points was changed during the study period Patients with exception points for other reasons were censored at the time of exception point approval, which mirrored the methodology in the study by Merion et al 9 Of note, some exceptions were approved at the discretion of the regional review board, but not based on the OPTN/UNOS policy 29 It would be difficult to assess an exact impact of the MELD-Na based allocation in patients with exception points, because of possible discrepancies of exception point practice among UNOS regions While the HCC exception point policy is national based, a decision on approval for exception point defers to regional review board Although this would be challenging, future studies will be necessary to review exception policies and their possible associations with the MELD and MELD-Na based allocation Limitations of this study include the use of the OPTN STAR registry, which is retrospective in nature and lacks detailed clinical information 30 The follow-up period in the MELD-Na group may be insufficient to draw firm conclusions regarding the impact of the new allocation system on post-transplant outcomes To confirm the impact of MELD-Na allocation on post-transplant outcomes, further investigations need to be conducted over longer periods of follow-up time In this study, to minimize a possible bias due to the discrepancy of follow-up

25 23 time between these two periods, all patients were censored at the end of each period, so that these two groups can be compared equally In addition, we set main end points as 90-day waitlist and one year post-transplant outcomes to reduce the bias We acknowledge that including patients and values from new registrations rather than using prevalent cohorts and longitudinal MELD/MELD-Na score entries may introduce bias in the analysis of entire cohort In conclusion, the new liver allocation based on MELD-Na score successfully decreased waitlist mortality Hyponatremic patients with low MELD scores received significant benefit from both MELD-Na based allocation and the Share 15 rule, whereas the positive impact was less pronounced in patients with higher scores The survival benefit of liver transplantation has shifted towards a higher score category in the MELD-Na period; therefore, liver allocation rules such as Share 15 and Share 35 need to be revised to fulfill the Final Rule under the MELD-Na based allocation 27

26 24 References 1 Kamath PS, Wiesner RH, Malinchoc M, et al A model to predict survival in patients with end-stage liver disease Hepatology 2001;33: Organ Procurement and Transplantation Network Liver and Intestine Committee, Allocation of Livers and Liver-Intestines, Biggins SW, Kim WR, Terrault NA, et al Evidence-based incorporation of serum sodium concentration into MELD Gastroenterology 2006;130: Bernardi M, Gitto S, Biselli M The MELD score in patients awaiting liver transplant: strengths and weaknesses J Hepatol 2011;54: Ruf AE, Kremers WK, Chavez LL, et al Addition of serum sodium into the MELD score predicts waiting list mortality better than MELD alone Liver Transpl 2005;11: Kim WR, Biggins SW, Kremers WK, et al Hyponatremia and mortality among patients on the liver-transplant waiting list N Engl J Med 2008;359: Feng S, O'Grady J Share 35: a liver in time saves lives? Am J Transplant 2015;15: Halazun KJ, Mathur AK, Rana AA, et al One Size Does Not Fit All--Regional Variation in the Impact of the Share 35 Liver Allocation Policy Am J Transplant 2016;16: Merion RM, Schaubel DE, Dykstra DM, et al The survival benefit of liver transplantation Am J Transplant 2005;5: Organ Procurement and Transplantation Network, Revised liver policy regarding HCC exception scores Organ Procurement and Transplantation Network, Organ Procurement and Transplantation Network, Upcoming MELD serum sodium policy implementation Organ Procurement and Transplantation Network, Merion RM, Wolfe RA, Dykstra DM, et al Longitudinal assessment of mortality risk among candidates for liver transplantation Liver Transpl 2003;9: Fisher RA, Heuman DM, Harper AM, et al Region 11 MELD Na exception prospective study Ann Hepatol 2012;11: Durand CM, Bowring MG, Thomas AG, et al The Drug Overdose Epidemic and Deceased-Donor Transplantation in the United States: A National Registry Study Ann Intern Med 2018;168:

27 25 15 Gonzalez SA, Trotter JF The rise of the opioid epidemic and hepatitis C-positive organs: A new era in liver transplantation Hepatology 2018;67: Kalra A, Wedd JP, Biggins SW Changing prioritization for transplantation: MELD-Na, hepatocellular carcinoma exceptions, and more Curr Opin Organ Transplant 2016;21: Terrault N, Berenguer M, Strasser S, et al International Liver Transplant Society Consensus Statement on HEPATITIS C MANAGEMENT IN LIVER TRANSPLANT RECIPIENTS Transplantation Herzer K, Gerken G When and How to Treat HCV Infection with the New Antivirals before or after Liver Transplantation Visc Med 2016;32: Flemming JA, Kim WR, Brosgart CL, et al Reduction in liver transplant wait-listing in the era of direct-acting antiviral therapy Hepatology 2017;65: Massie AB, Chow EK, Wickliffe CE, et al Early changes in liver distribution following implementation of Share 35 Am J Transplant 2015;15: Murken DR, Peng AW, Aufhauser DD, Jr, et al Same policy, different impact: Centerlevel effects of share 35 liver allocation Liver Transpl 2017;23: Sharma P, Schaubel DE, Goodrich NP, et al Serum sodium and survival benefit of liver transplantation Liver Transpl 2015;21: Nadim MK, DiNorcia J, Ji L, et al Inequity in organ allocation for patients awaiting liver transplantation: Rationale for uncapping the model for end-stage liver disease J Hepatol 2017;67: Orman ES, Ghabril M, Chalasani N Poor Performance Status Is Associated With Increased Mortality in Patients With Cirrhosis Clin Gastroenterol Hepatol 2016;14: e1 25 Ruebner R, Goldberg D, Abt PL, et al Risk of end-stage renal disease among liver transplant recipients with pretransplant renal dysfunction Am J Transplant 2012;12: Halazun KJ, Rana AA, Fortune B, et al No country for old livers? Examining and optimizing the utilization of elderly liver grafts Am J Transplant 2018;18:

28 26 27 Organ Procurement and Transplantation Network, Final Rule Organ Procurement and Transplantation Network, Organ Procurement and Transplantation Network, Board approves enhanced liver distribution system Organ Procurement and Transplantation Network, Organ Procurement and Transplantation Network, Liver and Intestine Committee, Guidance on MELD PELD exception review Organ Procurement and Transplantation Network, Massie AB, Kucirka LM, Segev DL Big data in organ transplantation: registries and administrative claims Am J Transplant 2014;14:

29 27 Figure legends Figure 1: a Score distribution at registration b Score distribution at transplant b Comparisons of serum sodium concentration at transplant according to score category Figure 2: Comparison of waitlist and post-transplant outcomes between the MELD and MELD-Na periods (Ref MELD period) (* P <005) a Hazard ratio for 90-day waitlist mortality in the MELD-Na period according to score categories b Hazard ratio for 90-day transplant probability in the MELD-Na period according to score categories c Hazard ratio for one year-liver graft loss in the MELD-Na period according to score range categories Figure 3: Comparison of cumulative incidence of waitlist mortality and transplant between the propensity score matched groups by period a Cumulative incidence of waitlist mortality and transplant b Standardized difference before and after matching The model achieved standardized difference less than 010 (10%) after the matching for all covariates Figure 4: Risk of hyponatremia for 90-day waitlist mortality in the MELD and MELD-Na periods (Ref normal sodium concentration [ mmol/l]) (* P <005) a Hazard ratio of mild hyponatremia ( mmol/l) in the MELD period b Hazard ratio of moderate hyponatremia ( mmol/l) in the MELD period c Hazard ratio of severe hyponatremia (<125 mmol/l) in the MELD period d Hazard ratio of mild hyponatremia ( mmol/l) in the MELD-Na period e Hazard ratio of moderate hyponatremia ( mmol/l) in the MELD-Na period

30 28 f Hazard ratio of severe hyponatremia (<125 mmol/l) in the MELD-Na period Figure 5: Comparison of mortality risk in the MELD-Na vs MELD periods Mortality risk by MELD/MELD-Na score for liver transplant recipients was compared to patients on the liver transplant waitlist The columns showed hazard ratios and p values by classifying the patients based on their final scores Columns highlighted in light blue, gray, and orange demonstrated not favoring transplant (staying on the waitlist), comparable, and favoring transplant The transition point justifying survival benefit of liver transplant shifted towards higher score category a Waitlist death includes mortality only b Waitlist death includes mortality and removal for too sick to transplant

31 29

32 Table 1 Patient characteristics in the MELD and MELD-Na periods Waitlist patient characteristics MELD period [n=18,850] MELD-Na period [n=14,512] P value Age at listing [year] 560 [49-62] 560 [48-62] 0863 Male 11,390 (604%) 8,651 (596%) 0134 Race White 13,520 (711%) 10,488 (723%) Black 1661 (88%) 1054 (73%) Hispanic 2,806 (149%) 2,280 (157%) <0001 Asian Other 585 (31%) 278 (15%) 453 (31%) 237 (16%) Primary liver disease Hepatitis C Non-alcoholic steatohepatitis 4,159 (221%) 3,225 (171%) 1,848 (127%) 3,041(210%) <0001 Alcohol Other 5,721 (304%) 5,745 (304%) 5,072 (350%) 4,551 (314%) Actual score at listing* 18 [14-26] 20 [15-28] <0001 Sodium at listing [mmol/l] 136 [ ] 137 [ ] 0213 Delta MELD 309 (SE 017) 33 (SE 019) 0009 Karnofsky <30% at listing 2,512 (136%) 1,790 (126%) 0007 Dialysis at listing 2,037 (108%) 1,531 (105%) 0464 Life support at listing 819 (43%) 540 (37%) 0006 Total time on waiting list [day] 93 [17-302] 62 [14-199] <0001 Follow-up time till 90 days [day] 90 [17-90] 62 [14-90] <0001 Waitlist status after 90 days Transplant 4,871 (258%) 4,666 (322%) Death/Removal for too sick 1,783 (95%) 1,051 (72%) <0001 Censored/Other removal reasons 12,196 (647%) 8,795 (606%) Transplant patient characteristics MELD period [n=6,547] MELD Na period [n=5,862] P value Recipient age at transplant [year] 56 [485-62] 56 [48-62] 0703 Recipient gender male 4191 (640%) 3,733 (637%) 0708 Actual score at transplant* 30 [23-37] 29 [22-35] <0001 MELD score at transplant 30 [23-37] 26 [19-346] <0001 MELD-Na score at transplant 315 [25-377] 29 [22-35] <0001 Sodium at transplant [mmol/l] 136 [ ] 135 [ ] <0001 Waiting time till transplant [day] 25 [7-94] 22 [7-70] <0001 Cold ischemia time [hour] 60 [48-75] 57 [45-71] <0001 Karnofsky <30% at transplant 2,121 (357%) 1,596 (278%) <0001 Pre-transplant dialysis 1,760 (269%) 1,333 (227%) <0001 Pre-transplant life support 993 (152%) 648 (112%) <0001 Donor age [year] 40 [27-53] 39 [27-53] 0783 Donor body mass index 266 [ ] 270 [ ] 0005 Donation after cardiac death 371 (57%) 391 (67%) 002 Donor location Local Regional National 3,812 (582%) 2,480 (379%) 255 (39%) 3,544 (605%) 2,045 (349%) 273 (47%) Distance to donor hospital [Mile] 88 [13-227] 86 [12-213] 0041 Length of hospital stay [day] 12 [8-20] 11 [7-18] <0001 Data were summarized using the median with interquartile range (IQR) for continuous variables and using percentage for discrete variables SE, standard error * MELD score in MELD period and MELD-Na score in MELD-Na period 0001

33 Table 2 Unadjusted waitlist and transplant mortality rates by score category in the MELD and MELD-Na periods MELD Death Waitlist outcome Waitlist outcome Post-transplant outcome [one year follow-up] Patient Years (PY) Death per 1000 PY Death + Too sick PY Death per 1000 PY Death PY Death per 1000 PY Total MELD-Na Total

34 Supplemental table 1 Comparison of waitlist patient characteristics between periods, within subcategories defined by each patient s calculated MELD and MELD-Na scores MELD period [n=181] MELD <15 and MELD-Na 15 MELD <35 and MELD-Na 35 MELD and MELD-Na 35 MELD-Na period [n=714] P value MELD period [n=137] MELD-Na period [n=281] P value MELD period [n=2,122] MELD-Na period [n=1,342] Age at listing [year] 58 [52-63] 58 [50-63] [48-61] 53 [44-60] [45-60] 53 [45-60] 0268 Male 128 (707%) 488 (627%) (650%) 179 (637%) (612%) 816 (608%) 083 Race White 142 (785%) 556 (779%) 108 (788%) 200 (712%) 1342 (632%) 863 (643%) Black 4 (22%) 27 (38%) 10 (73%) 20 (71%) 2828 (133%) 130 (97%) Hispanic 29 (160%) 107 (150%) 13 (95%) 50 (178%) 380 (179%) 268 (200%) 0012 Asian Others 4 (22%) 2 (11%) 17 (24%) 7 (10%) 4 (29%) 2 (15%) 7 (25%) 4 (14%) 87 (41%) 31 (15%) 53 (39%) 28 (21%) Primary diagnosis Hepatitis C Non-alcoholic steatohepatitis Alcohol Other 39 (215%) 40 (221%) 92 (129%) 176 (246%) < (234%) 14 (102%) 35 (125%) 40 (142%) (197%) 244 (115%) 122 (91%) 190 (142%) < (343%) 40 (221%) 266 (373%) 180 (252%) 59 (431%) 32 (234%) 142 (505%) 64 (228%) 818 (385%) 642 (303%) 636 (474%) 394 (294%) Body mass index 279 [ ] 274 [ ] [ ] 282 [ ] [ ] 290 [ ] 0522 Actual score at listing* 13 [13-14] 17 [16-18] < [33-34] 35 [35-36] < [37-42] 39 [37-42] 0014 MELD at listing 13 [13-14] 13 [12-14] [33-34] 341 [ ] < [37-42] 39 [37-427] 028 MELD-Na at listing 17 [16-19] 17 [16-18] [ ] 35 [35-36] < [372-42] 39 [37-42] 0338 Sodium at listing [mmol/l] 135 [ ] 134 [ ] [ ] 130 [ ] < [ ] 137 [ ] 0021 Karnofsky <30% at listing 3 (17%) 5 (07%) (341%) 111 (402%) (616%) 826 (626%) 0562 Dialysis at listing 0 (0%) 0 (0%) - 24 (175%) 56 (199%) (464%) 654 (487%) 0184 Life support at listing 0 (0%) 1 (01%) >09 9 (66%) 18 (65%) > (228%) 300 (225%) 09 Waitlist status after 90 days Transplant 14 (77%) 162 (227%) 92 (672%) 221 (786%) 1482 (698%) 963 (718%) < Death/Removal for too sick 17 (94%) 12 (17%) 31 (226%) 39 (139%) 524 (247%) 279 (208%) 0003 Censored/other 150 (829%) 540 (756%) 14 (102%) 21 (75%) 116 (55%) 101 (75%) Total time on waiting list [day] 91 [30-283] 94 [29-229] [4-21] 7 [3-15] [3-11] 5 [3-11] 0075 Follow-up time till 90 days [day] 90 [30-90] 90 [29-90] [4-21] 7 [3-15] [3-11] 5 [3-11] 0075 Waiting time until transplant**[day] 73 [49-125] 39 [17-84] [4-15] 6 [3-11] [3-8] 4 [2-8] 0007 * MELD score in MELD period and MELD-Na score in MELD-Na period ** In patients who received deceased donor liver transplant [Excluding patients who got an exception score] Data were summarized using the median with interquartile range (IQR) for continuous variables and using percentage for discrete variables P value