Gender disparity in liver transplant waiting-list mortality: The importance of kidney function

Authors

  • Ayse L. Mindikoglu,

    Corresponding author
    1. Division of Gastroenterology and Hepatology, Department of Medicine, University of Maryland School of Medicine, Baltimore, MD
    • Division of Gastroenterology and Hepatology, Department of Medicine, University of Maryland School of Medicine, 22 South Greene Street N3W50, Baltimore, MD 21201
    Search for more papers by this author
    • Telephone: 410-328-1358; FAX: 410-328-1897

  • Arie Regev,

    1. Division of Gastroenterology and Hepatology, Department of Medicine, Indiana University School of Medicine, Indianapolis, IN
    Search for more papers by this author
  • Stephen L. Seliger,

    1. Division of Nephrology, Department of Medicine, University of Maryland School of Medicine, Baltimore, MD
    Search for more papers by this author
  • Laurence S. Magder

    1. Division of Biostatistics and Bioinformatics, Department of Epidemiology and Preventive Medicine, University of Maryland School of Medicine, Baltimore, MD
    Search for more papers by this author

Errata

This article is corrected by:

  1. Errata: Erratum: Gender disparity in liver transplant waiting-list mortality: The importance of kidney function Volume 17, Issue 9, 1119, Article first published online: 22 August 2011

  • This work was supported in part by the Health Resources and Services Administration (contract 234-2005-370011C). 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. The project described was supported by Grant Number 1 K23 DK089008-01 from the National Institutes of Health (NIH) National Institute of Diabetes and Digestive and Kidney Diseases (to Ayse L. Mindikoglu M.D., M.P.H.) and its contents are solely the responsibility of the authors and do not necessarily represent the official views of the National Institute of Diabetes and Digestive and Kidney Diseases or the NIH.

Abstract

Previous studies of men and women on the liver transplantation (LT) waiting list, without taking transplantation rates into account, have suggested a higher risk of mortality for women on the waiting list. The objective of this study was to compare men and women with respect to dying within 3 years of registration on the LT waiting list and to take into account both the immediate mortality risks and the transplantation rates. The analysis was based on Organ Procurement and Transplantation Network data for patients with end-stage liver disease (ESLD) on the waiting list who were registered between February 2002 and August 2009. Competing risk survival analysis was performed to assess the gender disparity in waiting-list mortality; 42,322 patients and 610,762 person-months of waiting-list experience were included in the analysis. The risk of dying within 3 years of listing was 19% and 17% in women and men, respectively (P < 0.0001). Among patients with kidney disease and especially those not on dialysis with an estimated glomerular filtration rate (eGFR) ≥15 and <30 mL/minute/1.73 m2, women had a substantially higher risk of dying on the waiting list within 3 years of registration versus men (26% versus 20%, P = 0.001). This disparity was related to lower transplantation rates in women (transplantation rate ratio = 0.68, P < 0.0001). Controlling for eGFR and other variables related to mortality risk, we found that the overall female-male disparity disappeared. In conclusion, among patients with ESLD and kidney dysfunction who are not on dialysis, there is a substantial gender disparity in LT waiting-list mortality. Our analysis suggests as an explanation the fact that women have lower transplantation rates than men in this group. The lower transplantation rates can be explained in part by the fact that Model for End-Stage Liver Disease scores tend to be lower for women versus men because they are based on serum creatinine rather than the glomerular filtration rate. Liver Transpl 16:1147–1157, 2010. © 2010 AASLD.

Since February 27, 2002, the Model for End-Stage Liver Disease (MELD) score has been used to prioritize patients with end-stage liver disease (ESLD) awaiting liver transplantation (LT) in the United States.1-3 The MELD score is calculated with 3 laboratory parameters [the serum creatinine level, the prothrombin time international normalized ratio (INR), and the total bilirubin level] and predicts the 3-month mortality rate for patients with ESLD on the waiting list.1-3 An increase in any of the MELD parameters leads in turn to an increase in the MELD score and shortens the waiting time for LT because the MELD score is used to prioritize the sickest patients for liver allocation.1-3

Although the MELD-based allocation system has reduced the waiting time and mortality on the LT waiting list for patients with ESLD and has increased fairness in the liver allocation system, previous studies have shown that it might have induced a gender disparity.4, 5 Kanwal et al.4 reported that women were less likely to undergo LT than men. Moylan et al.5 showed that in the MELD era, women had lower transplantation rates and higher mortality on the United Network for Organ Sharing (UNOS) LT waiting list than men after they controlled for the MELD score and other covariates. In combination, lower transplantation rates in women and higher mortality rates on the LT waiting list would result in a substantial gender disparity in the risk of dying within 3 years of registration on the LT waiting list.

The risk of dying after registration on the LT waiting list depends on both the risk of dying while on the waiting list and the rate of transplantation. In this retrospective cohort study using data from the Organ Procurement and Transplantation Network (OPTN), we employed a competing risk analysis to compare men and women with respect to the likelihood of transplantation and the risk of death while they are on the LT waiting list.

Abbreviations:

BMI, body mass index; CG, Cockroft-Gault; CKD-EPI, Chronic Kidney Disease Epidemiology Collaboration; eGFR, estimated glomerular filtration rate; ESLD, end-stage liver disease; GFR, glomerular filtration rate; INR, international normalized ratio; LT, liver transplantation; MDRD, Modification of Diet for Renal Disease; MELD, Model for End-Stage Liver Disease; OPTN, Organ Procurement and Transplantation Network; UNOS, United Network for Organ Sharing.

PATIENTS AND METHODS

Study Population

The analysis was based on OPTN waiting-list data for 176,471 patients registered between October 1, 1987 and August 25, 2009. Patients were excluded for the following reasons: they underwent living donor LT (n = 3666); they were listed before February 27, 2002 (n = 94,704); they were younger than 18 years (n = 5743); they had a diagnosis other than ESLD (n = 17,530); they were exceptional cases (eg, patients with hepatocellular carcinoma who received a priority MELD score; n = 7003); they were removed from the waiting list because of multiple listing, an error, a transfer to another center, or other reasons (not specified in the database; n = 4809); or they were classified as status 1 (n = 694). Patients with ESLD who were listed for multiorgan transplantation and retransplantation were included in the analysis.

Variables

UNOS Standard Transplant Analysis and Research files were analyzed. The waiting-list variables used in the analysis were the patient code, waiting-list registration code, registration date, last follow-up date, days on the LT waiting list, gender, age, ethnicity, diagnosis, reason for removal from the waiting list, life support (ventilator, artificial liver, or other device), dialysis, serum total bilirubin level, serum creatinine level, prothrombin time INR, MELD score, exception status, transplant region, ABO blood type, diabetes, weight, and height.

In addition, the UNOS LT waiting-list history data set, which contained variables collected longitudinally from transplant candidates from the time of initial registration until removal from the waiting list, was used. This data set included repeated measures of serum creatinine, total bilirubin levels, INR and updates of the MELD score, allocation and dialysis status. Dialysis status was defined as dialysis twice in the week before the last serum creatinine measurement; these data were collected from every transplant candidate throughout his or her stay on the LT waiting list.

Using the information in the data set, we estimated the glomerular filtration rate (GFR) with the Modification of Diet for Renal Disease (MDRD) equation6, 7 and Chronic Kidney Disease Epidemiology Collaboration (CKD-EPI) equation,8 and we categorized the results according to the clinical practice guidelines for chronic kidney disease of the National Kidney Foundation Kidney Disease Outcomes Quality Initiative (stage 1, estimated GFR (eGFR) ≥90 mL/minute/1.73 m2 and not on dialysis; stage 2, eGFR ≥60 and <90 mL/minute/1.73 m2 and not on dialysis; stage 3, eGFR ≥30 and <60 mL/minute/1.73 m2 and not on dialysis; stage 4, eGFR ≥15 and <30 mL/minute/1.73 m2 and not on dialysis; and stage 5, eGFR <15 mL/minute/1.73 m2 or on dialysis).9-11 The body mass index (BMI) was categorized according to the World Health Organization classification (underweight, BMI <18.50 kg/m2; normal, BMI ≥18.50 and <25 kg/m2; overweight, BMI ≥25 and <30 kg/m2; and obese, BMI ≥30 kg/m2).12 BMI values <15 and >55 kg/m2 were categorized as missing because they were not considered biologically plausible.13

Statistical Analysis

Statistical analyses were performed with SAS 9.2 for Windows (SAS Institute, Inc., Cary, NC)14 and with the Criskcox SAS macro created by Kremers et al.15 The cumulative incidence plots and bar chart for transplantation rates were drawn with Minitab statistical software (Minitab, Inc., State College, PA).16

Reformatting the UNOS LT Waiting-List Data Set into a Person-Month Data Set

To facilitate the analysis, the LT waiting-list data set was restructured so that it consisted of 1 record for each month that each person spent on the waiting list. Each person-month record contained information on the most recent values of the prognostic variables corresponding to that month on the waiting list and an indication of whether the person died, underwent transplantation, was removed because of deteriorating or improved condition, or survived that month and stayed on the waiting list. Patients who were removed from the waiting list because of death, transplantation, or improved or deteriorating condition were classified as death, transplant, improved, and deteriorated, respectively, in that month. Those who were still on the waiting list at the end of study period were treated as right-censored.

The analysis was based on the waiting-list experience of the included patients from the initial listing to removal; this included periods when a patient was on the waiting list but was designated as temporarily inactive.

The percentages of missing estimated glomerular filtration rate (eGFR), MELD score, and dialysis values in the person-month file were 0.06%, 0.05%, and 0.4%, respectively. More than 99.5% of the person-months had no missing data and were included in the analysis.

Cumulative Incidence of Death and Transplantation on the LT Waiting List

We obtained unadjusted and adjusted estimates of the cumulative incidence of death and LT within 3 years of listing by fitting proportional hazards models to each cause of waiting-list removal and combining these results to estimate the cumulative incidence of each outcome. To do this, we used the Criskcox SAS macro written by Kremers et al.15 This is sometimes called a competing risk analysis because it includes multiple competing reasons for removal from the LT waiting list. In brief, this approach involves fitting separate Cox proportional hazards regression models to each cause of waiting-list removal. In our study, the competing causes were death, transplantation, removal from the waiting list due to deteriorating condition, and removal from the waiting list due to improved condition. These Cox models resulted in cause-specific hazard ratios showing the effect of each predictor on each cause of removal. The cumulative incidence of any specific outcome by a certain point in time was then calculated on the basis of the model-derived estimates of probabilities for that specific outcome and no competing outcome before the specified point in time. To assess the proportional hazards assumption with respect to gender, we introduced terms into the model corresponding to the interaction between gender and time.

Univariate and multivariate competing risk models were derived from the entire cohort and from subgroups defined by kidney function. The multivariate Cox models were based on time-varying values of the covariates and included age, ethnicity, etiology of ESLD, region, blood type, diabetes, MELD score, and dialysis. In addition, we fit a model with the same predictors but replaced the MELD score with time-varying values of the logarithm of INR, logarithm of the bilirubin level, and eGFR. The quantitative predictors (age, MELD, eGFR, logarithm of INR, and logarithm of the bilirubin level) were treated as linearly related to the logarithm of hazards. In the model, eGFR was truncated at 100 mL/minute/1.73 m2 because exploratory work suggested that the hazards did not change after that. Also, the model allowed a different effect of eGFR for those on dialysis. In the multivariate models, cumulative incidence estimates applied only to a specified subgroup defined by specific values for all variables in the model. Therefore, to obtain insight into the effect of gender on cumulative incidence in the multivariate models, we compared males and females with respect to cumulative incidence in a subgroup with values for quantitative covariates that corresponded to a moderate risk of transplantation and death and with modal values for categorical variables in the model.

We assessed the proportional hazards assumptions with respect to gender in these models by including an interaction term between time and gender. In general, there was no strong evidence against the proportional hazards assumption with respect to hazards for death. However, there was some evidence that the female/male transplantation rate ratio varied over time. For example, in the unadjusted model (shown later in Table 2, first row), when we allowed the female/male transplantation rate ratio to change over time, it was estimated to be 0.74 at the initial listing but 0.84 after 2 years on the waiting list. However, the resulting cumulative incidence estimates for this (and other models) remained virtually unchanged when the proportional hazards assumption was relaxed; therefore, for the sake of simplicity, we report the original models.

To calculate P values for observed differences in the cumulative incidence of death, we used the following approach. The data set was randomly divided into 4 subsets. The difference between men and women with respect to the 3-year cumulative incidence of death was estimated for each subset. The sample standard deviation of these 4 difference estimates was calculated, and this provided us an unbiased estimate of the standard error of difference estimates based on a sample that was one-fourth the size of our actual sample. To obtain an unbiased estimate of the standard error of our actual estimates based on the full sample, we divided this value by 2 (because the standard deviation of estimates from a sample is half the standard deviation of estimates found in a sample one-fourth of the size). This process was repeated 30 times, and the average standard error found in these calculations was used to calculate P values for cumulative incidence differences (reported later in Tables 2 and 3).

RESULTS

After excluding the patients who did not fulfill the study inclusion criteria, we included a total of 42,322 patients with ESLD and 610,762 person-months in the analysis.

Characteristics of the Patients on the LT Waiting List

Table 1 shows the distribution of demographic, clinical, and laboratory characteristics of patients with ESLD at waiting-list registration. Among 42,322 patients, 35% were women, and 65% were men. The etiologies of ESLD differed between men and women listed for LT. Although men were more likely to have ESLD secondary to alcohol, primary sclerosing cholangitis, or viral etiologies, women were more likely to have ESLD secondary to most other etiologies. The median MELD score was lower in women versus men both at listing (15 versus 16, P < 0.0001; Table 1) and at the last follow-up on the LT waiting list (18 versus 19, P < 0.0001). The median serum creatinine concentration was lower in women versus men at listing (0.9 versus 1.1 mg/dL, P < 0.0001) and at the last follow-up on the LT waiting list (1.1 versus 1.2 mg/dL, P < 0.0001). However, the median eGFR estimated by MDRD was substantially reduced in women versus men at listing (68.6 versus 80.2 mL/minute/1.73 m2, P < 0.0001; Table 1) and at the last follow-up (60.2 versus 70.9 mL/minute/1.73 m2, P < 0.0001), respectively. Similar results were obtained when the CKD-EPI equation was used to estimate the GFR (Table 1). Although eGFR in women was lower than that in men, the proportion of men who were on dialysis was significantly higher than the proportion of women (5% versus 4%, P = 0.002; Table 1).

Table 1. Demographic and Clinical Characteristics of 42,322 Patients with ESLD on the LT Waiting List Who Were Registered Between February 27, 2002 and August 25, 2009
Patient Characteristic*Women (n = 14,840)Men (n = 27,482)P Value
  • *

    Quantitative variables are reported as medians and interquartile ranges, except for age, which is reported as means and standard deviations.

  • Cirrhosis secondary to drug/industrial exposure, histiocytosis, sarcoidosis, granulomatous disease, hemachromatosis, alpha-1 antitrypsin deficiency, Wilson's disease, and cystic fibrosis.

  • The percentages of patients who were temporarily inactive at the last follow-up visit were as follows: 29% were still on the waiting list; 45% died; 0% underwent transplantation; and 7% and 20% were removed from the LT waiting list because of improved or deteriorating condition, respectively.

Age (years)53.6 (9.9)52.6 (8.8)<0.0001
Ethnicity [n (%)]  <0.0001
 White10,438 (70)20,630 (75) 
 Black1363 (9)1945 (7) 
 Hispanic2403 (16)3877 (14) 
 Asian449 (3)773 (3) 
 Others187 (1)257 (1) 
Etiology of cirrhosis [n (%)]  <0.0001
 Hepatitis C5427 (37)13,286 (48) 
 Hepatitis B250 (2)1035 (4) 
 Alcohol1967 (13)6480 (24) 
 Cryptogenic1961 (13)2372 (9) 
 Nonalcoholic fatty liver disease1265 (9)1163 (4) 
 Primary sclerosing cholangitis703 (5)1625 (6) 
 Primary biliary cirrhosis1643 (11)241 (1) 
 Autoimmune hepatitis1154 (8)400 (1) 
 Other470 (3)880 (3) 
MELD score15 (10)16 (10)<0.0001
Total bilirubin level (mg/dL)2.6 (4.2)2.6 (3.7)0.0003
Serum creatinine level (mg/dL)0.9 (0.6)1.1 (0.7)<0.0001
INR1.4 (0.5)1.4 (0.6)<0.0001
Serum sodium level [n (%)]  <0.0001
 <125 mmol/L212 (2)436 (2) 
 125 to <135 mmol/L2810 (28)5780 (31) 
 135 to <145 mmol/L6827 (68)12,071 (65) 
 ≥145 mmol/L239 (2)390 (2) 
 Missing data13,549  
eGFRMDRD (mL/minute/1.73 m2)68.6 (47.1)80.2 (51.8)<0.0001
eGFRMDRD [n (%)]  <0.0001
 eGFR ≥90 mL/minute/1.73 m2, not on dialysis (stage 1)4261 (29)10,621 (39) 
 eGFR ≥60 and <90 mL/minute/1.73 m2, not on dialysis (stage 2)4656 (31)8287 (30) 
 eGFR ≥30 and <60 mL/minute/1.73 m2, not on dialysis (stage 3)3937 (27)5321 (19) 
 eGFR ≥15 and <30 mL/minute/1.73 m2, not on dialysis (stage 4)1006 (7)1364 (5) 
 eGFR <15 mL/minute/1.73 m2 or on dialysis (stage 5)980 (7)1889 (7) 
eGFRCKD-EPI (mL/minute/1.73 m2)71 (50.6)80.9 (48.5)<0.0001
eGFRCKD-EPI [n (%)]  <0.0001
 eGFR ≥90 mL/minute/1.73 m2, not on dialysis (stage 1)4633 (31)10,866 (40) 
 eGFR ≥60 and <90 mL/minute/1.73 m2, not on dialysis (stage 2)4474 (30)8000 (29) 
 eGFR ≥30 and <60 mL/minute/1.73 m2, not on dialysis (stage 3)3721 (25)5254 (19) 
 eGFR ≥15 and <30 mL/minute/1.73 m2, not on dialysis (stage 4)1012 (7)1407 (5) 
 eGFR <15 mL/minute/1.73 m2 or on dialysis, not on dialysis (stage 5)1000 (7)1955 (7) 
Dialysis in the week before listing [n (%)]  0.002
 Yes666 (4)1425 (5) 
 No14,174 (96)26,057 (95) 
Life support [n (%)]  0.22
 Yes318 (2)541 (2) 
 No14,515 (98)26,938 (98) 
 Missing data10  
Region [n (%)]  <0.0001
 1495 (3)1132 (4) 
 21498 (10)3296 (12) 
 31882 (13)3638 (13) 
 41973 (13)2861 (10) 
 52759 (19)4839 (18) 
 6414 (3)829 (3) 
 71343 (9)2359 (9) 
 81002 (7)1758 (6) 
 91161 (8)2198 (8) 
 101219 (8)2109 (8) 
 111094 (7)2463 (9) 
ABO blood type [n (%)]  0.02
 A5413 (36)10,369 (38) 
 B1829 (12)3352 (12) 
 AB555 (4)1100 (4) 
 O7043 (47)12,661 (46) 
Diabetes [n (%)]  0.15
 Yes3609 (24)6513 (24) 
 No11,231 (76)20,969 (76) 
BMI [n (%)]  <0.0001
 <18.50 kg/m2 (underweight)351 (2)282 (1) 
 ≥18.50 to <25 kg/m2 (normal)4556 (31)6918 (25) 
 ≥25 to <30 kg/m2 (overweight)4401 (30)10,421 (38) 
 ≥30 kg/m2 (obese)5416 (37)9714 (35) 
 Missing data263 
Reason for removal from the LT waiting list [n (%)]  <0.0001
 Still waiting4312 (29)6861 (25) 
 Death2665 (18)4386 (16) 
 Transplantation6326 (43)13,893 (51) 
 Condition improved474 (3)615 (2) 
 Condition deteriorated1063 (7)1727 (6) 

Cumulative Incidence of Death and Transplantation

Tables 2 and 3 show the unadjusted and adjusted cumulative incidence of death within 3 years of listing for LT. Although the differences in the cumulative incidence of transplantation, death, and removal due to deterioration between men and women were small during the first months after listing on the LT waiting list, the difference became larger and stayed relatively constant after 3 years (Fig. 1). Therefore, we based our analysis on 3 years of follow-up after patients were listed for LT.

Figure 1.

Cumulative incidence of death, LT, and removal from the LT waiting list due to deterioration (A,C,E) in the entire cohort (controlled for age, ethnicity, etiology, region, blood type, diabetes, dialysis, and MELD score) and (B,D,F) in the subgroup of patients not on dialysis with eGFR values ≥15 and <30 mL/minute/1.73 m2 (stage 4 kidney disease) (controlled for age, ethnicity, etiology, region, blood type, diabetes, and MELD score). These risk estimates apply specifically to those with mean or mode values of the predictors (ie, age of 54 years, hepatitis C infection, blood type O, white ethnicity, no diabetes, and region 5). The MELD score was set to 20 to obtain moderate estimates of death and transplantation. [Color figure can be viewed in the online issue, which is available at wileyonlinelibrary.com.]

Table 2. Comparison of Women and Men with Respect to the Cumulative Incidence of Death Within 3 Years of Registration on the LT Waiting List
 Female/Male Mortality Ratio on the LT Waiting ListP ValueFemale/Male Transplant Ratio on the LT Waiting ListP ValueCumulative Incidence of Death Within 3 Years of Listing
WomenMenDifference95% Confidence IntervalP Value
  1. NOTE: The GFR was estimated with the MDRD equation.

Entire cohort1.001.0000.77<0.00010.190.170.020.011-0.025<0.0001
eGFR ≥90 mL/minute/1.73 m2, not on dialysis (stage 1)0.870.0020.78<0.00010.150.16−0.01−0.016 to 0.0040.256
eGFR ≥60 mL/minute/1.73 m2, not on dialysis (stage 2)0.900.0180.68<0.00010.170.160.010.001-0.0210.035
eGFR ≥30 mL/minute/1.73 m2, not on dialysis (stage 3)0.920.1070.64<0.00010.210.180.030.012-0.0520.002
eGFR ≥15 mL/minute/1.73 m2, not on dialysis (stage 4)1.020.8190.68<0.00010.260.200.060.022-0.0860.001
eGFR <15 mL/minute/1.73 m2 or on dialysis (stage 5)1.200.0240.900.0490.260.220.040.006-0.0710.021
Creatinine level ≥1.0 and ≤4.0 mg/dL, not on dialysis1.23<0.00010.86<0.00010.210.170.040.031-0.054<0.0001
Table 3. Comparison of Women and Men with Respect to the Cumulative Incidence of Death Within 3 Years of Registration on the LT Waiting List Based on Models That Adjust for Differences in Confounding Variables
 Female/Male Mortality Ratio on the LT Waiting ListP ValueFemale/Male Transplant Ratio on the LT Waiting ListP ValueCumulative Incidence of Death Within 3 Years of Listing*
WomenMenDifference95% Confidence IntervalP Value
  • NOTE: The GFR was estimated with the MDRD equation.

  • *

    These risk estimates apply specifically to those with mean or mode values of the predictors (ie, age of 54 years, hepatitis C infection, blood type O, white ethnicity, no diabetes, and region 5). The MELD score was set to 20, the total bilirubin level was set to 4.7 mg/dL, and the INR was set to 1.6 to obtain moderate estimates of death and transplantation.

  • Adjusted for age, ethnicity, etiology, region, blood type, diabetes, dialysis, and MELD score.

  • Adjusted for age, ethnicity, etiology, region, blood type, diabetes, dialysis, logarithm of the bilirubin level, logarithm of INR, and eGFR.

  • §

    Adjusted for age, ethnicity, etiology, region, blood type, diabetes, and MELD score.

Entire cohort1.15<0.00010.82<0.00010.310.260.050.032-0.057<0.0001
Entire cohort0.920.00130.71<0.00010.210.210−0.005 to 0.0160.273
eGFR ≥90 mL/minute/1.73 m2, not on dialysis (stage 1)§1.060.2540.83<0.00010.340.320.02−0.003 to 0.0520.080
eGFR ≥60 mL/minute/1.73 m2 and GFR <90 mL/minute/1.73 m2, not on dialysis (stage 2)§1.100.0430.78<0.00010.330.290.040.021-0.063<0.0001
eGFR ≥30 mL/minute/1.73 m2 and GFR <60 mL/minute/1.73 m2, not on dialysis (stage 3)§1.24<0.00010.80<0.00010.270.210.060.036-0.078<0.0001
eGFR ≥15 mL/minute/1.73 m2 and GFR <30 mL/minute/1.73 m2, not on dialysis (stage 4)§1.430.00030.77<0.00010.180.120.060.029-0.0960.0003
eGFR <15 mL/minute/1.73 m2 or on dialysis (stage 5)§1.180.0580.860.00510.180.150.03−0.002 to 0.0610.067
Serum creatinine level ≥1.0 and ≤4.0 mg/dL, not on dialysis§1.21<0.00010.81<0.00010.290.240.050.036-0.072<0.0001

In the entire cohort, women and men had similar rates of mortality on the LT waiting list, but women had lower transplantation rates (Table 2). Combining these competing events, we estimated that 19% of women died on the LT waiting list within 3 years of registration versus only 17% of men (P < 0.0001; Table 2). The disparity in 3-year mortality was greater when we considered subsets of patients with kidney disease, and it was greatest among those with stage 4 kidney disease (26% versus 20%, P = 0.001; Table 2). This disparity was due to the substantially lower transplantation rates in women versus men in this group (female/male transplantation rate ratio = 0.68, P < 0.0001; Table 2).

With adjustments for age, ethnicity, etiology, region, blood type, diabetes, dialysis, and MELD score, women had a 5% higher 3-year cumulative incidence of death versus men in the moderate-risk subgroup (Table 3 and Fig. 1A; 3-year cumulative incidence of 31% for women versus 26% for men, P < 0.0001). The gender disparity in mortality was substantially higher in the group not on dialysis with eGFR values ≥15 and <30 mL/minute/1.73 m2 (stage 4 kidney disease; 18% versus 12%, P = 0.0003; Table 3 and Fig. 1B) because after adjustments for age, ethnicity, etiology, region, blood type, diabetes, and MELD score, women had substantially higher rates of mortality on the waiting list (female/male mortality rate ratio = 1.43, P = 0.0003; Table 3). The disparity in mortality was also high among patients not on dialysis with eGFR values ≥30 and <60 mL/minute/1.73 m2 (stage 3 kidney disease; 3-year cumulative incidence of 27% for women versus 21% for men, P < 0.0001; female/male mortality rate ratio = 1.24, P < 0.0001) and in the group not on dialysis with serum creatinine concentrations ≥1.0 and ≤4.0 mg/dL (3-year cumulative incidence of 29% for women versus 24% for men, P < 0.0001; female/male mortality rate ratio = 1.21, P < 0.0001; Table 3).

The female excess in mortality disappeared when eGFR was added to the model adjusted for age, ethnicity, etiology, region, blood type, diabetes, dialysis, bilirubin, and INR (Table 3, second row). Controlling for eGFR, we found that men on the waiting list had higher mortality rates and higher transplantation rates. The higher mortality rates and higher transplantation rates in men offset each other, and the 3-year cumulative incidence of death became similar in men and women (Table 3). In the same model that included age, ethnicity, etiology, region, blood type, diabetes, dialysis, bilirubin, INR, and eGFR, any 30-U decrease in eGFR (estimated with the MDRD equation) increased the risk of death on the waiting list by a factor of 0.5 among patients who were not on dialysis (regression coefficient = −0.02499, hazard ratio = 0.975, P < 0.0001).

To assess whether the lower transplantation rates in women versus men could be completely explained by MELD scores, we compared their transplantation rates for each MELD score separately. We found that, even within groups defined by MELD scores, women generally had lower transplantation rates (Fig. 2).

Figure 2.

LT rates per person-year in women and men by the MELD score.

DISCUSSION

Our analysis showed that among patients with ESLD and kidney disease who were not on dialysis, women had a substantially higher mortality risk on the LT waiting list. This difference appears to be primarily due to lower transplantation rates in women, especially those with eGFR values ≥15 and <30 mL/minute/1.73 m2 (stage 4 kidney disease). These findings might be explained by the fact that among those not on dialysis with eGFR values ≥15 and <30 mL/minute/1.73 m2 (stage 4 kidney disease), MELD scores for women tend to be lower than MELD scores for men because the MELD score is based on the serum creatinine concentration rather than GFR, and for a given eGFR level, the serum creatinine concentration is lower for women versus men on account of lower average creatinine generation in women. Lower MELD scores result in lower transplantation rates. The disparity in transplantation rates is less among those with eGFR values <15 or on dialysis (stage 5 kidney disease) because any transplant candidate on dialysis or with a serum creatinine concentration greater than 4.0 mg/dL is given a value of the serum creatinine concentration equal to 4.0 mg/dL for the calculation of the MELD score. In multivariate models that control for the MELD score (ie, models that base the estimates of female/male mortality rate ratios on those with equal MELD scores), the disparity in transplantation is less pronounced. However, among those with kidney disease, when we controlled for MELD and other variables, women had a substantially higher immediate risk of dying on the waiting list, and this is consistent with the fact that among men and women with the same MELD scores, women tend to have lower GFR values.

We had hypothesized that the lower transplantation rates for women versus men were due to lower MELD scores in women. However, the lower transplantation rates for women versus men were not entirely explained by lower MELD scores for women because even within groups defined by MELD scores, women generally have lower transplantation rates (Fig. 2). The reason for this disparity is unknown; but it might be due to the fact that women tend to be smaller than men and might not be candidates because of recipient-donor liver size mismatch.17

We found that women with ESLD had significantly lower eGFR values than men on the LT waiting list. Although the explanation for this is uncertain, a possibility is the fact that the decision to list a patient on the LT waiting list is influenced by serum creatinine concentrations unadjusted for gender in MELD scores. Our recent study showed that among patients with ESLD and eGFR values <30 mL/minute/1.73 m2 who were not on dialysis and underwent LT, women were less likely to undergo simultaneous liver-kidney transplantation than men (odds ratio = 0.5, P < 0.0001).18 These results may reflect the dominant influence of serum creatinine rather than GFR in decision making even in simultaneous liver-kidney transplantation because all subjects in the study underwent LT, and lower transplantation rates due to possible recipient-donor liver size mismatch cannot be considered a factor for lower simultaneous liver-kidney transplantation rates in women. Our results showed that although eGFR values in women were lower than those in men, the proportion of men who underwent dialysis was significantly higher than the proportion of women. These results suggest that men and woman do not have equal access to dialysis because of underestimation of kidney dysfunction in women. Kausz et al.19 showed that the initiation of dialysis treatment for kidney failure was significantly delayed in women versus men and suggested that one of the reasons for this delay was the underestimation of renal dysfunction in women. In one study, women had significantly lower mean GFR values versus men before the initiation of dialysis, although they had lower mean serum creatinine concentrations.20

Our analysis differs in several distinct aspects from previously published studies showing gender disparity in LT. Using competing risk methods, we estimated differences in mortality between men and women that took into consideration both differences in immediate mortality rates among those remaining on the LT waiting list and differences in transplantation rates. Also, to the best of our knowledge, this is the first and largest retrospective cohort study evaluating the survival rates by gender in patients with ESLD on the LT waiting list in the context of time-varying values of the MELD score as well as renal parameters such as the serum creatinine concentration, dialysis status, and GFR estimated with both the MDRD and CKD-EPI equations. Transplant candidates might stay on the waiting list for months to years, and fluctuation in their disease severity may change the values of the laboratory tests at listing. In addition, the dialysis status might change between the initial listing and last follow-up. Our analysis took all the repeated measures of candidates' laboratory tests (ie, the serum creatinine level, sodium level, total bilirubin level, and INR) and the MELD scores throughout their stay on the LT waiting list into account and captured any change in the dialysis status; in contrast, in previous studies, the analysis was based on the values of predictor variables and the dialysis status at either listing or last follow-up.

Another difference from other studies that have investigated gender disparity on the LT waiting list is that we included in our analysis the waiting-list time during which patients were temporarily inactive. UNOS provided longitudinal information on these patients, and we saw no reason not to include them. In fact, leaving them out would have biased estimates of waiting-list mortality downward because almost half of the waiting-list deaths occurred among patients who were temporarily inactive.

Our study has certain limitations. First, because we used a retrospective cohort, we were not able to assess the accuracy of the patients' eGFR values and evaluate whether there was any significant difference in comparison with the measured GFR values of women and men. A few reports have indicated that GFR might be overestimated when creatinine-based equations are used in patients with cirrhosis. Therefore, in some patients in our cohort, the GFR may have been overestimated by creatinine-based estimating equations. In a small group of patients with cirrhosis, Cholongitas et al.21 reported only moderate correlations between measured GFR values and GFR values estimated with the MDRD equation and between measured GFR values and creatinine clearance estimated with the Cockroft-Gault (CG) equation (r = 0.64 and r = 0.55, respectively). Gonwa et al.22 showed that the MDRD equation was better for estimating renal function in patients with cirrhosis than other equations, but it was still not as precise as when it was used to estimate renal function in other populations. Skluzacek et al.23 reported that in patients with cirrhosis, the MDRD and CG equations overestimated the GFR and creatinine clearance, respectively. Greater precision was reported when cystatin C–based equations were used to estimate the GFR in patients with cirrhosis instead of creatinine-based equations such as the MDRD equation.24 Several investigators have confirmed that cystatin C is a better indicator for renal dysfunction in patients with cirrhosis versus the serum creatinine concentration.24-30 However, these studies have limitations in design and methods. Prospective studies are needed for the validation of novel biomarkers of kidney function (eg, cystatin C) in cirrhosis to demonstrate their superiority in accuracy and precision versus serum creatinine. We believe that large prospective studies using gold-standard GFR measurement methods are also needed to validate existing creatinine-based and cystatin C–based GFR-estimating equations in subjects with cirrhosis before the implementation of the eGFR into the MELD equation is considered.7, 8, 31-33

In our study cohort, serum creatinine assays were performed in many different laboratories using different creatinine calibrations, and this might have caused systematic error in the GFR estimation. In order to prevent such errors in GFR values estimated with the MDRD and CKD-EPI equations, laboratories in the United States should perform isotope dilution mass spectrometry traceable creatinine calibration as recommended by the National Kidney Disease Education Program.34 Because our cohort consisted of patients from different transplant centers in the United States and serum creatinine concentrations were measured in different laboratories, we were not able to determine the methods used to measure the serum creatinine concentration in each patient. This limitation might cause bias not only in GFR estimations but also in MELD score calculations.35

Lastly, the cause of death and the etiology of kidney disease were not available in the database.

Our results suggest that the use of the unadjusted serum creatinine concentration as an indicator of kidney function in the MELD score results in a gender disparity. They stress the need to implement more accurate methods for the determination of kidney function in place of serum creatinine in the MELD score to provide equitable liver allocation for patients on the LT waiting list. On the basis of these findings, drug dose adjustments and decisions regarding intravenous contrast administration in patients with cirrhosis should be based on the GFR and not solely on the serum creatinine concentration to prevent drug overdose, drug-related toxicity, and intravenous contrast nephropathy. The possibility of underestimation of kidney dysfunction with creatinine-based methods (eg, the MDRD, CKD-EPI, and CG equations and 24-hour urinary creatinine clearance) should always be taken into account to ensure the safest care for patients with cirrhosis.

In conclusion, among patients not on dialysis with ESLD and kidney dysfunction, women had lower transplantation rates, which resulted in a higher risk of dying on the waiting list. These differences were likely caused in part by underestimation of renal dysfunction in women when the serum creatinine concentration, rather than the GFR, was used to calculate the MELD scores.

Acknowledgements

The authors thank Walter K. Kremers, Ph.D. (Mayo Clinic Transplant Center), for providing the Criskcox SAS macro for the competing risk analysis and Charles Howell, M.D. (Division of Gastroenterology and Hepatology, Department of Medicine, University of Maryland School of Medicine) for his valuable input in the manuscript.

Ancillary

Advertisement