The impact of serum sodium concentration on mortality after liver transplantation: A cohort multicenter study

Authors

  • Muhammad F. Dawwas,

    Corresponding author
    1. Hepatobiliary and Liver Transplant Unit, Addenbrooke's Hospital, Cambridge University Hospitals National Health Service (NHS) Foundation Trust, Cambridge, UK
    2. Clinical Effectiveness Unit, The Royal College of Surgeons of England, London, UK
    • Hepatobiliary and Liver Transplant Unit, Box 210, Addenbrooke's Hospital, Hills Road, Cambridge, CB2 2QQ, UK
    Search for more papers by this author
    • Telephone: 44 1223 2245151; FAX: 44 1223 216111

  • James D. Lewsey,

    1. Clinical Effectiveness Unit, The Royal College of Surgeons of England, London, UK
    2. Health Services Research Unit, Department of Public Health and Policy, London School of Hygiene and Tropical Medicine, London, UK
    Search for more papers by this author
  • James M. Neuberger,

    1. Liver Unit, Queen Elizabeth Hospital, Birmingham, UK
    Search for more papers by this author
  • Alexander E. Gimson

    1. Hepatobiliary and Liver Transplant Unit, Addenbrooke's Hospital, Cambridge University Hospitals National Health Service (NHS) Foundation Trust, Cambridge, UK
    Search for more papers by this author

  • See Editorial on Page 1080

Abstract

Modification of the current allocation system for donor livers in the United States to incorporate recipient serum sodium concentration ([Na]) has recently been proposed. However, the impact of this parameter on posttransplantation mortality has not been previously examined in a large risk-adjusted analysis. We assessed the effect of recipient [Na] on the survival of all adults with chronic liver disease who received a first single organ liver transplant in the UK and Ireland during the period March 1, 1994 to March 31, 2005 (n = 5,152) at 3 years, during the first 90 days, and beyond the first 90 days, adjusting for a wide range of recipient, donor, and graft characteristics. Compared to those with normal [Na] (135–145 meq/L; n = 3,066), severely hyponatremic recipients ([Na] <130 meq/L, n = 541), had a higher risk-adjusted mortality at 3 years (hazard ratio [HR] 1.28; 95% confidence interval [CI], 1.04–1.59; P < 0.02). The excess mortality was, however, confined to the first 90 days (HR 1.55; 95% CI, 1.18–2.04; P < 0.002) with no significant difference thereafter. This was also true for hypernatremic recipients ([Na] >145 meq/L, n = 81), who had an even greater risk-adjusted mortality compared to normonatremic recipients (overall: HR 1.85; 95% CI, 1.25–2.73; P < 0.002; ≤90 days: HR 2.29; 95% CI, 1.42–3.70; P < 0.001; >90 days: HR 1.12; 95% CI, 0.55–2.29; P = 0.8), whereas mildly hyponatremic recipients ([Na] 130–134 meq/L, n = 1,127) had similar risk-adjusted mortality to those with normal [Na] at the same time points. In conclusion, recipient [Na] is an independent predictor of death following liver transplantation. Attempts to correct the [Na] toward the normal reference range are an important aspect of pretransplantation management. Liver Transpl 13:1115–1124, 2007. © 2007 AASLD.

In recent years there has been increasing debate concerning the most appropriate system for allocation of organs for liver transplantation.1, 2 In the United States, the introduction of the Model for End-Stage Liver Disease (MELD) in 2002 was predicated on a desire to prioritize organs to those cases with the highest short-term risk of mortality while waiting for a graft.3 The MELD score as a basis for organ allocation has many advantages, including the utilization of objective parameters—serum creatinine, bilirubin, and international normalized ratio of prothrombin time—not easily subject to bias or manipulation. Since the introduction of the MELD-based prioritization scheme, both waiting list mortality and waiting times have fallen in the United States and there has been no reduction in transplant survival.4 Further attempts to refine this score have included assessment of changes in the score over time5, 6 and the possible inclusion of other parameters such as serum sodium concentration [Na], presence of persistent ascites, and hepatic venous pressure gradient.7–10 The recent demonstration that recipient [Na] improves the ability of the MELD score to predict mortality among patients with end-stage chronic liver disease7–9 has led to proposals to incorporate it in the current MELD-based allocation system of donor livers in the United States.7, 9, 11

In contrast to the pretransplantation setting, outcome after transplantation is dependent on a number of factors, which include both donor quality and recipient fitness. MELD score has little impact on posttransplantation survival except at the extreme high range and thus an allocation process that prioritizes cases with a higher MELD score will be expected to maximize individual transplant benefit.12 The objective of this cohort multicenter study is to examine the impact of recipient serum sodium concentration [Na] on mortality after liver transplantation.

Abbreviations

[Na], serum sodium concentration; MELD, Model for End-Stage Liver Disease; HR, hazard ratio; CPM, central pontine myelinolysis; CI, confidence interval.

PATIENTS AND METHODS

We used data that were prospectively collected and validated for the UK and Ireland Liver Transplant Audit. Details of this database and evidence of its completeness, accuracy, and reliability have been described elsewhere.13 All patients with chronic liver disease who received a first transplant between March 1, 1994 and March 31, 2005 in both countries were eligible for inclusion in the study. We excluded patients if they were younger than 16 years or underwent a multiorgan transplant (i.e., simultaneous transplantation of a liver and another organ).

In the UK and Ireland, patients with end-stage chronic liver disease are currently selected for transplantation at 1 of the 8 regional liver transplant centers if they are deemed to have a poor prognosis or quality of life without transplantation and 5-year posttransplantation survival prospects of greater than 50%.14 Once selected for transplantation, patients are prioritized on the waiting list and allocated a donor organ as deemed appropriate by their transplant center.

We collected data on a wide range of clinically plausible recipient, donor and graft risk factors. All recipient laboratory data, including serum sodium [Na], were measured immediately prior to transplantation. In line with previous studies, the range of serum recipient [Na] was categorized as: severe hyponatremia ([Na] < 130 meq/L),8, 9, 15–22 mild hyponatremia ([Na] = 130–134 meq/L),9, 21 normal ([Na] = 135–145 meq/L), and hypernatremia ([Na] >145 meq/L). The MELD score was calculated using serum creatinine, serum total bilirubin, and international normalized ratio of prothrombin time values using the formula: MELD score = 10 × [0.957 × Loge(creatinine mg/dL) + 0. 378 × Loge(bilirubin mg/dL) + 1.120 × Loge(international normalized ratio) + 0.643]. The serum creatinine concentration was set to 4.0 mg/dL for those who required renal support and had lower values. No upper or lower limits were applied to the score and no extra points were awarded to hepatocellular carcinoma recipients and other special cases. A high MELD score corresponds to a poorer prognosis in the absence of liver transplantation.3 MELD scores were stratified into the following categories as defined by Merion et al.12 (≤11, 12–14, 15–17, 18–20, 21–29, 30–39, and ≥40). We categorized the body mass index according to the World Health Organization guidelines as follows: nonobese (18.5–24.9 kg/m2), overweight (25.0–29.9 kg/m2), obese (≥30.0 kg/m2), or underweight (<18.5 kg/m2). All other numerical data were used as continuous variables in the multivariable analyses.

We adopted a 9-category chronic liver disease classification system (Table 1) similar to that used by Roberts et al.23 Patients were classified according to both diagnostic code and, if present, free-text diagnosis. To ensure that, in the event of multiple diagnoses, patients were assigned to the diagnosis that was deemed most likely to influence their prognosis, disease classification was undertaken in a hierarchical order: cancer, hepatitis C cirrhosis, and primary sclerosing cholangitis. For example, patients with a coded diagnosis of hepatitis C cirrhosis and a free-text diagnosis of hepatocellular carcinoma were assigned to the “cancer” category. All patients with Wilson's disease and Budd-Chiari syndrome were assigned to the metabolic and other liver diseases categories, respectively, regardless of their disease onset. Hepatic encephalopathy was defined using the West Haven criteria.24, 25 Recipient functional status was defined using the modified Eastern Cooperative Oncology Group performance status score.26, 27

Table 1. Distribution of Clinical Characteristics at Time of Transplantation by Recipient Serum Sodium Concentration [Na] Category in the UK and Ireland Between March 1, 1994 and March 31, 2005
CharacteristicOverallSerum sodium concentration [Na] meq/LP-value
<130130–134135–145>145
  1. NOTE: Values are means [standard deviation] unless indicated as percentages or P-values.

  2. Abbreviations: HBV, hepatitis B virus; BMI, body mass index; CMV, cytomegalovirus.

Number of recipients5,1525411,1273,06681 
% of recipients100.010.521.959.51.6 
Recipient diagnosis (%)     <0.001
 Primary biliary cirrhosis17.612.917.017.620.9 
 Primary sclerosing cholangitis9.86.78.210.911.1 
 Alcoholic liver disease17.522.919.415.811.1 
 Autoimmune and cryptogenic cirrhosis disease12.415.512.711.813.6 
 HCV cirrhosis15.118.917.014.113.6 
 HBV cirrhosis4.66.76.43.12.5 
 Cancers13.46.811.516.49.9 
 Metabolic liver disease5.46.34.95.77.4 
 Other liver diseases4.23.32.94.69.9 
Recipient age (years)50.4 [11.4]50.2 [10.4]51.5 [10.3]50.2 [11.8]48.9 [13.2]<0.005
Recipient gender (% male)59.661.661.959.254.30.2
Recipient race (% nonwhite)13.916.214.413.219.80.1
Calculated MELD score17.3 [7.9]22.3 [9.3]18.4 [7.4]15.8 [7.1]21.1 [10.5]<0.0001
Albumin (gm/dL)3.0 [0.6]2.8 [0.6]2.9 [0.6]3.2 [0.6]2.8 [0.9]<0.0001
Recipient BMI (kg/m2)25.5 [4.9]25.4 [5.1]25.9 [4.7]25.6 [5.0]24.7 [3.8]0.1
Preoperative renal support (%)7.211.76.76.521.0<0.001
Preoperative ventilation (%)1.61.91.41.217.3<0.001
Poor preoperative functional status (%)19.138.122.813.140.7<0.001
Previous upper abdominal surgery (%)19.916.819.220.019.80.4
Diuretic therapy (%)60.778.076.352.449.4<0.001
Clinically detectable ascites (%)57.983.371.148.258.0<0.001
Grade III/IV encephalopathy (%)2.55.02.11.612.4<0.001
Graft type (%)     0.03
 Whole94.995.896.294.291.4 
 Segmental5.14.33.85.88.6 
Cold ischemic time (minutes)666.8 [188.6]675.2 [195.9]664.5 [192.9]660.9 [184.7]644.6 [204.2]0.4
Organ appearance (% suboptimal)16.320.815.916.712.80.06
Donor age (years)42.8 [14.9]43.8 [14.6]43.6 [15.1]42.6 [15.0]40.7 [14.8]0.05
Donor type (%)     <0.04
 Cadaveric heart-beating98.598.898.998.296.3 
 Living0.40.60.40.42.5 
 Cadaveric non-heart-beating1.10.60.71.41.2 
Donor cause of death (%)     0.5
 Trauma21.122.919.720.618.5 
 Other78.977.180.379.481.5 
Donor BMI (kg/m2)24.9 [4.3]25.0 [4.1]25.1 [4.6]24.9 [4.3]24.5 [4.0]0.3
Donor/recipient gender match (%)     0.7
 Male/male37.838.839.437.332.1 
 Female/male21.822.722.521.922.2 
 Male/female16.215.614.216.521.0 
 Female/female24.222.923.924.324.7 
Donor/recipient blood group match (%)     <0.004
 Identical93.289.592.294.090.1 
 Compatible6.39.47.45.68.6 
 Incompatible0.51.10.40.41.3 
Donor/recipient CMV serology match (%)     <0.02
 Positive match65.766.670.164.173.8 
 Positive mismatch16.817.014.717.316.4 
 Negative match17.516.415.218.69.8 

We used survival analysis to analyze mortality after liver transplantation. Patients lost to follow-up were censored at the date of last follow-up. Patients requiring further transplantation were not censored at the time of retransplantation. Unadjusted mortality estimates were calculated using the Kaplan-Meier method. Differences in these estimates between different categories of patients based on their [Na] were compared using the log-rank test for trend.

Unadjusted and adjusted overall mortality estimates at 3 year as well as within and beyond the first 90 days were calculated using proportional hazards regression. The hazard ratios (HRs) thus generated indicate the relative risk of death of a given category of patients based on their [Na] at the time of transplantation compared to those with normal concentration. Only those clinically plausible recipient, donor, recipient, and graft risk factors that were well-recorded in the database with nonmissing values of over 85% were included in the multivariable models. Those risk factors were: liver disease category (1–9), recipient age, race, calculated MELD score, serum albumin, body mass index, requirement for preoperative renal support, requirement for preoperative ventilation, presence of poor preoperative functional status (defined as a score of 4 or 5 on the Eastern Cooperative Oncology Group performance status score), history of previous upper abdominal surgery, use of diuretic therapy, presence of clinically detectable ascites, presence of grade III/IV encephalopathy, graft type (categorized as whole or segmental, including split and reduced grafts), organ cold ischemic time and appearance (categorized by the implanting surgeon as either healthy or suboptimal, as a result of steatosis or other reasons, based on the gross appearance of the liver), donor age, type (categorized as cadaveric heart-beating, live, or cadaveric non-heart-beating), cause of death (categorized as trauma or other) and body mass index, donor/ recipient gender match, donor/recipient blood group match, donor/ recipient cytomegalovirus serology match, and year of transplantation. We also included transplant center in all models as a stratifying variable.

Implausible values of body mass index (<10 kg/m2 or >100 kg/m2), cold ischemic time (>40 hours), serum bilirubin (<0.1 mg/dL), serum creatinine (<0.1 mg/dL or >15 mg/dL), and serum albumin (<0.7 gm/dL or >6.0 gm/dL) were considered to be missing.

To ensure that patients with missing values were not excluded from the multivariable analysis, we used the technique of multiple imputation as described by Royston.28, 29 Missing values were imputed 10 times using switching regression to create 10 complete datasets that were each independently analyzed. Analyses were then pooled to give final estimates. To assess whether the impact of serum [Na] on posttransplantation mortality might be different at different values of other covariates in the risk model, the presence of a statistically significant interaction between each [Na] category and each of the following risk factors was individually tested: liver disease category, calculated MELD score, presence of clinically detectable ascites, presence of grade III/IV hepatic encephalopathy, use of diuretic therapy, requirement for preoperative renal support, and requirement for preoperative mechanical ventilation.

The frequency of postoperative complications within the first 90 days among survivors of this period was also assessed. Postoperative renal complications were defined as renal dysfunction requiring short-term or long-term renal support. Postoperative sepsis was defined as the occurrence of microbiologically-proven evidence of infection in blood, wound, sputum, ascitic/drain fluid, urine or other site requiring antibiotic therapy. Postoperative vascular complications were defined as the occurrence of hepatic artery thrombosis, portal vein thrombosis, inferior vena cava/hepatic vein occlusion, or hemorrhage requiring reoperation. Postoperative biliary complications were defined as the occurrence of a biliary tract leak or stricture, either of which required endoscopic, percutaneous, or surgical intervention. Biliary complications that did not require such intervention were not scored. The incidence of acute rejection was calculated including only those episodes that were both histologically confirmed and required additional immunosuppressive therapy. Causes of death within the first 90 days were categorized into the following groups: cardiothoracic, sepsis and multiorgan failure, donor organ failure/other graft-related cause, cerebrovascular accident, malignancy, and other causes.

Differences in postoperative resource utilization (as measured by the perioperative blood product requirements and time to discharge from the hospital), complications, and causes of death across serum [Na] categories were compared using the χ2 test and Kruskal-Wallis, for categorical and continuous variables, respectively. These comparisons were also carried out individually between each [Na] category and the normonatremic group. Differences in recipient clinical characteristics were compared using the χ2 test and analysis of variance as appropriate. A P-value of 0.05 or less was considered to indicate a statistically significant result. All analyses were performed using Stata version 9 (StataCorp, College Station, TX).

RESULTS

Clinical Characteristics

Between March 1, 1994 and March 31, 2005, 5,152 adults with chronic liver disease underwent a first single organ liver transplant at 8 centers in the UK and Ireland. Two additional recipients were excluded from the analysis because of missing survival data. The prevalence of hyponatremia at transplantation in the study population was 32.4% (this being severe in 10.5% of recipients) whereas that of hypernatremia was considerably lower at 1.6%. The results of [Na] were unknown in 337 (6.5%) recipients. Compared to recipients with normal [Na] at transplantation (Table 1), severely hyponatremic recipients were more often grade III/IV encephalopathic, more likely to exhibit clinically detectable ascites, receive diuretic therapy, and require transplantation for alcoholic liver disease and HCV cirrhosis. They also had poorer functional status at transplantation, higher ventilation and renal support requirements, greater MELD scores, and were more likely to receive a suboptimal-looking graft. The age, gender, and racial distribution of hyponatremic and normonatremic groups was similar. Hypernatremic recipients also had clinical and biochemical evidence of more advanced liver disease, with substantially higher requirements for renal support and ventilation compared to those with serum [Na] but, unlike severely hyponatremic patients, were more likely to be transplanted for primary biliary cirrhosis and “other” liver diseases.

Unadjusted Mortality

The overall unadjusted 3-year mortality (Table 2 and Fig. 1) of severely hyponatremic recipients was higher than that of those with normal [Na] (HR 1.50; 95% confidence interval [CI], 1.25–1.82; P < 0.001). The greatest impact of severe hyponatremia on posttransplant mortality was observed during the first 90 days (HR 1.95; 95% CI, 1.52–2.51; P < 0.001) with no significant effect thereafter (HR 1.15; 95% CI, 0.87–1.51; P = 0.3). Hypernatremic recipients also fared worse at 3 years than those with normal [Na] (HR 2.27; 95% CI, 1.57–3.30; P < 0.001), their mortality being particularly high during the first 90 days (HR 3.50; 95% CI, 2.25–5.46; P < 0.001) but no different from that of those with normal [Na] among survivors of the first 90 days posttransplantation (HR 1.20; 95% CI, 0.60–2.42; P = 0.6). In contrast, the mortality of mildly hyponatremic recipients was similar to that of those with normal [Na] at all time-points during the first 3 years (overall: HR 1.14; 95% CI, 0.97–1.35; P = 0.1; ≤90 days: HR 1.23; 95% CI, 0.97–1.56; P = 0.09; >90 days: HR 1.08; 95% CI, 0.87–1.34; P = 0.5). The mortality of patients with an unknown [Na] was not statistically different from that of those for whom this data was available (90 days: 11.6%, 95% CI, 8.60–15.5 vs. 9.2%, 95% CI, 8.5–10.1; 1 yr: 16.4%, 95% CI, 12.8–20.8 vs. 14.8%, 95% CI, 13.8–15.8; 3 years: 22.3%, 95% CI, 18.2–27.2 vs. 21.3%, 95% CI, 20.1–22.6; P = 0.9).

Table 2. Kaplan-Meier 90-day, 1-year, and 3-year Mortality (and 95% Confidence Intervals) by Recipient Serum Sodium Concentration [Na] Category in the UK and Ireland Between March 1, 1994 and March 31, 2005
[Na] Category (meq/L)90-day1-year3-year
<13015.0 (12.2–18.3)21.5 (18.2–25.2)27.4 (23.7–31.6)
130–1349.3 (7.8–11.2)15.0 (13.0–17.2)21.9 (19.5–24.6)
135–1457.7 (6.9–8.8)13.0 (11.9–14.3)19.6 (18.2–21.1)
>14526.0 (17.8–37.0)32.7 (23.6–44.3)37.5 (27.7–49.5)
Figure 1.

Kaplan-Meier survival graphs for liver transplant recipients in the UK and Ireland between March 1, 1994 and March 31, 2005 by recipient serum sodium concentration [Na] category (log-rank test for trend: P < 0.001).

Adjusted Mortality

The overall risk-adjusted mortality (Fig. 2) of severely hyponatremic recipients was higher than that of normonatremic recipients at 3 years (HR 1.28; 95%CI, 1.04–1.59; P < 0.02). The excess mortality was, however, confined to the first 90 days posttransplantation (HR 1.55; 95% CI, 1.18–2.04; P < 0.002) with no significant difference thereafter (HR 1.05; 1.10 95% CI; 0.78–1.42, P = 0.7). This was also true for hypernatremic recipients (overall: HR 1.85; 95% CI, 1.25–2.73; P < 0.002; ≤90 days: HR 2.29; 95% CI, 1.42–3.70; P < 0.001; >90 days: HR 1.12; 95% CI; 0.55–2.29; P = 0.8) whereas mildly hyponatremic recipients had similar risk-adjusted mortality at the above time-points (overall: HR 1.07; 95% CI, 0.90–1.27; P = 0.4; ≤90 days: HR 1.11; 95% CI, 0.86–1.42; P = 0.4; >90 days: HR 1.04, 95% CI; 0.83–1.31; P = 0.7). Higher MELD scores were not independently predictive of higher mortality at any time point after liver transplantation (P > 0.05) except among recipients with MELD score ≥40 (representing only 2.1% of the cohort) whose mortality during the first 90 days was significantly higher than those with MELD scores ≤11 (HR 1.91; 95% CI, 1.07–3.43; P < 0.03). In addition, the impact of serum recipient [Na] on posttransplantation mortality was not modified by other relevant risk factors since no statistically significant interactions were noted between any of the sodium [Na] categories and liver disease category, calculated MELD score, presence of clinically detectable ascites, presence of grade III/IV hepatic encephalopathy, use of diuretic therapy, requirement for preoperative renal support, or requirement for preoperative mechanical ventilation (Table 3).

Figure 2.

Unadjusted and adjusted HRs (and 95% CI) for overall mortality at 3 years, during the first 90 days, and beyond the first 90 days in liver transplant recipients in the UK and Ireland between March 1, 1994 and March 31, 2005 by recipient [Na] category compared with those with normal [Na] ([Na] = 135–145 meq/L).

Table 3. Adjusted Hazard Ratios (and 95% Confidence Intervals) for Mortality During the First 90 Days by Interaction Term of Recipient Serum Sodium Concentration [Na] Category With: Liver Disease Category, Calculated MELD Score, Presence of Clinically Detectable Ascites, Presence of Grade III/IV Hepatic Encephalopathy, Use of Diuretic therapy, Requirement for Preoperative Renal Support and Requirement for Preoperative Mechanical Ventilation Among Liver Transplant Recipients in the UK and Ireland Between March 1, 1994 and March 31, 2005
InteractionSerum sodium concentration [Na] meq/L
<130P value130–134P value>145P value
  1. Abbreviations: PBC, primary biliary cirrhosis; PSC, primary sclerosing cholangitis; ALD, alcoholic liver disease; AID, autoimmune and cryptogenic disease; HCV, hepatitis C virus; HBV, hepatitis B virus.

Diagnosis (reference PBC)      
 PSC1.20 (0.39–3.75)0.70.96 (0.35–2.65)0.91.92e–18 (0)1.0
 ALD1.01 (0.40–2.52)1.00.87 (0.38–1.97)0.71.93 (0.31–12.05)0.5
 AID1.20 (0.47–3.06)0.71.47 (0.66–3.27)0.32.53 (0.43–15.00)0.3
 HCV cirrhosis0.56 (0.22–1.47)0.20.54 (0.23–1.28)0.22.38 (0.37–15.33)0.4
 HBV cirrhosis0.45 (0.09–2.12)0.31.18 (0.38–3.67)0.86.74 (0.41–111.55)0.2
 Cancer1.09 (0.35–3.39)0.90.80 (0.33–1.95)0.63.25 (0.42–24.98)0.3
 Metabolic0.56 (0.15–2.03)0.41.02 (0.35–2.95)1.01.09 (0.09–13.20)0.9
 Other0.73 (0.20–2.69)0.60.63 (0.18–2.24)0.51.64 (0.24–11.38)0.6
MELD1.00 (0.97–1.02)0.80.99 (0.97–1.02)0.61.00 (0.96–1.03)0.8
Ascites1.20 (0.59–2.44)0.61.60 (0.92–2.78)0.092.40 (0.65–8.86)0.2
Grade III/IV encephalopathy0.83 (0.33–2.08)0.70.70 (0.21–2.34)0.60.43 (0.13–1.38)0.2
Diuretic therapy0.94 (0.51–1.74)0.81.18 (0.69–2.00)0.51.64 (0.63–4.28)0.3
Renal support0.80 (0.37–1.71)0.61.12 (0.53–2.37)0.81.02 (0.36–2.87)1.0
Ventilation0.64 (0.17–2.50)0.50.53 (0.14–1.94)0.30.61 (0.20–1.85)0.4

Postoperative Complications, Resource Utilization, and Causes of Death

Compared to those with normal [Na] (Table 4), severely hyponatremic recipients who survived the first 90 days posttransplantation were more likely to develop renal (P < 0.001), septic (P < 0.001), and vascular (P < 0.05) postoperative complications, were more often functionally incapacitated at 90 days (P < 0.001) but were less likely to experience acute graft rejection (P < 0.004). In addition, severely hyponatremic recipients who survived the first 90 days required considerably higher volumes of perioperative blood products (Table 5) and longer median hospital stay (22 days vs. 18 days; P < 0.0001).

Table 4. Frequency of Postoperative Complications (%) by Serum Recipient Sodium Concentration [Na] Category Among Survivors of the First 90 Days Following Liver Transplantation in the UK and Ireland Between March 1, 1994 and March 31, 2005
ComplicationOverall N = 5,152Serum sodium concentration [Na] meq/LP value
<130130–134135–145>145
Renal dysfunction (requiring short- or long-term renal support)9.219.710.37.612.5<0.001
Sepsis48.656.952.246.368.4<0.001
Vascular complications9.011.59.18.77.00.3
Biliary complications8.77.09.08.710.50.6
Acute rejection34.629.432.336.626.8<0.004
Poor functional status5.88.48.24.711.1<0.001
Table 5. Resource Utilization By Recipient Serum Sodium Concentration [Na] Category Among Survivors of the First 90 Days Following Liver Transplantation in the UK and Ireland Between March 1, 1994 and March 31, 2005
ParameterOverall N = 4,572Serum sodium concentration [Na] meq/LP-value
<130130–134135–145>145
  • NOTE: Values are medians; values in parentheses indicate interquartile range unless indicated as percentages or P-values.

  • *

    During and within the first 48 hours of transplantation.

Perioperative* blood product requirements (units)      
 Blood6 (7)9 (8)7 (8)5 (7)7 (9)<0.0001
 Platelets4 (10)5 (14)5 (14)3 (10)8 (14)<0.0001
 Fresh frozen plasma10 (11)12 (11)11 (11)9 (10)12 (9)<0.0001
 Cryoprecipitate0 (8)0 (8)0 (8)0 (2)0.5 (8)<0.0001
Time to discharge (days)18 (15)22 (18)19 (16)18 (14)19 (18)<0.0001

The frequency of postoperative complications and resource utilization of hypernatremic recipients were also generally greater than those with normal [Na] (Tables 4 and 5).

No statistically significant differences in the frequency of recorded causes of death (Table 6) were noted across the different sodium [Na] categories (P = 0.6).

Table 6. Causes of Death (%) in the First 90 Days by Recipient Serum Sodium Concentration [Na] Category Following Liver Transplantation in the UK and Ireland Between March 1, 1994 and March 31, 2005
Cause of death categoryOverall N = 483Serum sodium concentration [Na] meq/L
<130130–134135–145>145
Cardio-thoracic17.011.117.117.714.3
Donor organ failure/other graft-related cause12.28.615.211.419.1
Sepsis/multi-organ failure54.764.254.353.657.0
Cerebrovascular accident2.53.71.02.54.8
Malignancy0.61.31.000
Other13.011.111.414.84.8

DISCUSSION

We have shown that recipient serum sodium concentration measured immediately prior to liver transplantation is an independent predictor of mortality following the procedure.

Although the detrimental impact of both severe recipient hyponatremia and hypernatremia on posttransplantation survival was confined to the first 90 days, they also had deleterious effects on the frequency of postoperative complications, functional status, and resource utilization, even among those who survived this period.

Our study has some limitations that need to be acknowledged. First, although [Na] is an objective, easily reproducible biochemical parameter, it is one that is known to be influenced by factors such as diuretic therapy and dialysis. We believe, however, that this possibility is unlikely to have affected our results since both of these factors were adjusted for in the risk models and no statistically significant interactions between any of the sodium [Na] categories and those factors were evident in the multivariable analyses. However, given the high proportion of hypernatremic patients who received preoperative renal support, the possibility that these patients' true [Na] levels had in fact been in the hyponatremic range before they were artificially elevated by renal support cannot be ruled out.

Second, the fact that the prevalence of severe hyponatremia in our study population was less than half that reported in previous studies of hospitalized cirrhotic patients15, 18, 19, 22 implies a significant degree of selection bias by transplant clinicians with regard to candidacy for liver transplantation. However, this, if anything, is likely to have minimized, rather than amplified, the observed impact of [Na] on posttransplantation mortality in our study.

To our knowledge, only a handful of previous studies, most of which describing small single center experiences, have demonstrated the presence of a univariate association between recipient hyponatremia and higher posttransplantation mortality.30–34 On multivariable analysis, however, all but one of those studies did not find this association to be independent of other risk factors30–32, 34 whereas the only one that did33 was likely underpowered, with too few postoperative deaths relative to the number of risk factors considered.35

For a number of reasons, we believe our finding that severely hyponatremic liver transplant recipients fare worse than those with normal [Na] is both clinically and physiologically plausible.

First, hyponatremia has long been shown to predict mortality among hospitalized patients irrespective of their underlying disease.36–40

Second, [Na] is a component of the Acute Physiology and Chronic Health Evaluation (APACHE) score as well as the Physiological and Operative Severity Score for the enUmeration of Mortality and Morbidity (POSSUM), both of which have been previously shown to be useful in predicting surgical outcome.41, 42

Third, hyponatremia, being the biochemical footprint of an activated neurohumoral cascade,43 may represent a more inclusive and accurate measure of severity of liver disease and overall fitness for transplantation than classical liver biochemistry. Previous studies have shown it to be significantly associated with a number of features of advanced liver disease such as reduced liver size,19 persistent ascites,8, 11, 18, 19, 22, 43 hepatic encephalopathy,22 hemodynamic dysfunction,16–19, 43–45 and high bilirubin levels.8, 18 Hyponatremia may also be precipitated by life-threatening events such as sepsis18, 19, 22 and variceal bleeding19 whereas its development during the course of liver disease is an important step in the pathway leading to hepatorenal syndrome.16, 22, 43

Fourth, a strong association between hyponatremia and the development of central pontine myelinolysis (CPM) is well-recognized.46 Liver transplant recipients have a CPM incidence of 10 to 30% at autopsy47–50 and constitute the third largest group of patients, after alcoholics and those with chronic hyponatremia, in whom CPM has been described, accounting for 17% of all reported cases since 1986.51 The etiology of CPM in these patients is not fully understood but is thought to develop as a result of large perioperative osmolar shifts in predisposed individuals by virtue of chronic hyponatremia, malnutrition, debility, and, possibly, immunosuppression.48, 49, 51–57 This contention is supported by our observation of significantly higher perioperative blood product requirements among hyponatremic recipients, who were shown by others to have a CPM incidence of 17.4 to 57.1%.48, 49, 54, 57 Furthermore, most previous studies of CPM after liver transplantation have reported a preponderance of either preoperative hyponatraemia48, 49, 54, 57–59 and/or significant swings in recipient [Na] levels with either an increase or decrease of >15 meq/L in the 24 to 72 hours immediately preceding the onset of symptoms eventually attributed to CPM.47, 60 This could also account for the higher mortality we observed in those with hypernatremia, a much less prevalent biochemical disorder in patients with chronic liver disease.22 Nevertheless, the impact of severe hyponatremia on the subsequent development of CPM as a cause of death could not unfortunately be verified from this cohort as the incidence of CPM was not prospectively monitored in our database.

Our study has implications for the liver transplant community and physicians managing patients with liver disease. The reduced survival following liver transplantation in the severe hyponatremia group strongly suggests the importance of efforts to correct this prior to surgery. Managing hyponatremia by cessation of diuretics, intravascular volume expansion to reduce nonosmolar elevation of antidiuretic hormone release and increase free-water clearance, as well as hemofiltration61 may be necessary. The advent of vasopressin receptor antagonists, which have recently been shown to normalize the serum [Na] of hyponatremic cirrhotic patients,20, 21, 62 promise further improvements in the management of these patients in the future. However, whether the correction of serum [Na] levels into the normal range will improve the transplant outcome for such cases will require further study.

During the perioperative period, the aim of management should be to prevent major fluctuations in the [Na], rather than correct it, and therefore extreme caution needs to be exercised while administering sodium-rich fluids (particularly blood products) and sodium bicarbonate intraoperatively.49, 54–56 However, in the context of high volume requirement, coagulopathy, and poor renal function, maintaining static [Na] levels can be a challenging task, and in extreme situations, the intraoperative utilization of hemofiltration/dialysis has been advocated.53, 55

It is of note that hypernatremic recipients, who comprised only 1.6% of our cohort, had the lowest posttransplantation survival of all the [Na] groups. This observation has not been noted elsewhere. Although hypernatremia has previously been described in association with variceal bleeding,63 increased insensible water losses,64 and the use of lactulose65 and cholestyramine therapy,66 its pathogenesis and optimal management in the context of chronic liver disease remain poorly defined and further studies are therefore required to examine these issues.

Recent suggestions that the liver organ allocation system in the United States should be modified to include serum [Na] along with MELD,7, 9, 11 thus potentially prioritizing donor livers to severely hyponatremic candidates, may impact on liver transplantation outcomes. As has been shown previously,13, 67, 68 we observed that MELD was poorly predictive of posttransplantation mortality, a finding that could explain the observation that the implementation of the MELD-based allocation system in the United States in 2002 and consequent prioritization of donor livers to the sickest recipients has not resulted in worse transplant outcomes.4 Although severe hyponatremia has a deleterious impact on both waiting list and posttransplantation mortality (as this study demonstrates), its impact on the former has been shown to be substantially greater.7–9, 11 Therefore, a liver allocation scheme prioritizing organs to hyponatremic recipients might be likely to maximize survival benefits from the procedure.

In conclusion, we found that both severe recipient hyponatremia and hypernatremia are associated with higher mortality following liver transplantation, both with and without risk adjustment. Efforts to correct the [Na] prior to surgery in such cases need to be carefully assessed.

Acknowledgements

We thank the following members of the UK and Ireland Liver Transplant Audit and their departments: Professor Nigel Heaton, Dr. John O'Grady, Mr. Mohamed Rela, and Susan Landymore (King's College Hospital, London, UK); Mr. David Mayer and Bridget Gunson (Queen Elizabeth Hospital, Birmingham, UK); Mr. Stephen Pollard, Mr. Raj Prasad, Mr. Giles Toogood, and Olive McGowan (St James' Hospital, Leeds, UK); Mr. Neville Jamieson and Claire Jenkins (Addenbrooke's Hospital, Cambridge, UK); Dr Andrew Bathgate, Mr. John Forsythe, and Karen Tuck (The Royal Infirmary of Edinburgh, Edinburgh, UK); Mr. Keith Rolles, Professor Andrew Burroughs, and Dr. Nancy Rolando (Royal Free Hospital, London, UK); Mr. Derek Manas, Mr. Bryon Jaques, and Liesl Smith (Freeman Hospital, Newcastle, UK); Professor Oscar Traynor and Mr. Emir Hoti (St. Vincent's Hospital, Dublin, Republic of Ireland); Dr. David Collett and Kerri Barber (UK Transplant, Bristol, UK); and Dr. Jan van der Meulen and Lynn Copley (the Royal College of Surgeons of England, London, UK).

Ancillary