Female Liver Transplant Recipients with the Same GFR as Male Recipients Have Lower MELD Scores—A Systematic Bias

Authors


* Corresponding author: Andrew K. Burroughs, Andrew.Burroughs@royalfree.nhs.uk

Abstract

Women have lower glomerular filtration (GFR) than men for the same serum creatinine (Cr) value, not accounted for in model for end-stage liver disease (MELD). We compare male/female Cr, GFR (using MDRD formula) and respective MELD scores in 403 Cr measurements using standard (sCr), O'Leary modified (mCr) and Compensated (cCr) Jaffe and Enzymatic (eCr) in 158 liver disease patients, mCr in 208 liver transplantation (LT) candidates, and EDTA-Cr51-GFR in 38 other candidates for LT; considering each female as male, a ‘corrected’ Cr was derived. MELD scores were calculated for measured and ‘corrected’ Cr in females. Median Cr and GFR in females were lower than males (p < 0.05). Both MDRD and EDTA-Cr51 GFR were lower in females than males, despite lower Cr values. In females, each MELD score was lower than the corresponding MELD–corrected Cr (p < 0.001) with ≥three-point difference in liver disease patients: 25%[sCr]; 23%[mCr]; 11%[eCr]; and 14%[cCr]. In 65% of female LT candidates, two- or three-point difference was found. Females with liver disease have lower GFR than males for the same Cr value; correcting Cr increases MELD score by two or three points in 65% of female LT candidates. MELD score adjustment in females would ensure equal LT priority by gender.

Abbreviations: 
MELD

Model for End-stage Liver Disease

Bil

Bilirubin

Cr

creatinine

eGFR

estimated GFR

sCr

standard kinetic Jaffe

mCr

O'Leary modified Jaffe

eCr

enzymatic

cCr

compensated kinetic Jaffe.

Introduction

The model for end-stage liver disease (MELD) is used to prioritize candidates for liver transplantation (LT) in the United States (1). The MELD score is based on objective and widely available laboratory tests derived by multivariable analysis: serum bilirubin (Bil), serum creatinine (Cr) and the international normalized ratio of prothrombin time (INR) (2,3).

However, MELD is not without limitations. Significant variations of MELD score have been found due to different laboratory methodologies for INR measurement (4,5). Utilization of different laboratory assays for Cr gives rise to significantly different MELD scores, which could lead to inequalities in prioritization of candidates, especially in those with the highest priority for LT (more jaundiced and with higher MELD score)(6). Despite this, the major advantage of MELD is the evaluation of renal function, which is an established marker of prognosis in cirrhosis, and is a surrogate marker of renal function, i.e. a correlate or glomerular filtration rate (GFR) (7,8).

Apart from difficulties in measurements, it is well recognized that Cr concentration is influenced by several factors unrelated to renal function such as total muscle mass. The latter can lead to discrepancies in Cr concentration between individuals with the same renal function but of different age, race and sex (9). Thus, individuals may have an over- or underestimation of their renal function and GFR, which is the best marker of renal function, when the estimation is based only on serum Cr. Elderly patients may have an overestimation of their renal function compared to younger individuals, and South Asian patients compared to Afro-Caribbean ones (10,11). Similarly, women have lower Cr compared to men with the same renal function (same GFR) (9).

The difference between females and males in terms of renal function (GFR) based on serum Cr measurement is not taken into account in the MELD score. Thus, women may not be receiving equal priority for LT if for the same serum Cr concentration, they have worse renal function, an adverse prognostic factor, but have the same MELD score. This could represent a systematic discrimination against women. An important premise for this hypothesis is that Cr within the MELD score is completely or almost completely a surrogate for renal function alone. The aim of our study was (a) to compare Cr and estimated GFR (eGFR) values between males and females in patients with liver disease (using four different assays for Cr) and in candidates for LT (using a single method for Cr measurement), and (b) to evaluate the influence of gender on GFR measurement and on MELD score. We also evaluated a separate cohort of transplant candidates using EDTA-Cr51 measurement for ‘true’ GFR to evaluate if this GFR differed in the same way between males and females, similar to eGFR.

Patients and Methods

We assessed 403 consecutive blood samples obtained prospectively during routine clinical care from 158 outpatients or inpatients with abnormal liver function tests, seen at the Royal Free Hospital (RFH). The samples were stored at –20°C prior to analysis. ‘Real’ Cr concentration was determined on each sample using four different methods, as described in detail in a previous paper (6):

  • 1Standard kinetic Jaffe (sCr) method
  • 2O'Leary modified Jaffe (mCr),
  • 3Enzymatic creatinine (eCr),
  • 4Compensated (rate blanked) kinetic Jaffe (cCr).

The above four methods are among the most widely used for estimating Cr in hospital laboratories in the UK and the USA: the kinetic Jaffe (or one of its modifications) in 75% (10% the O'Leary modified Jaffe) and 73%, respectively, and the enzymatic method in 7% and 27%, respectively (12).

For each Cr method, GFR was calculated using the modification of diet in renal disease (MDRD) formula, which is considered superior for estimating GFR (eGFR) from Cr in adults (13,14), and in cirrhotic patients (15,16) compared to other equations:

image

The equation does not require weight because the results are reported normalized to 1.732 body surface area, which is an accepted average adult surface area. The latter is an advantage in cirrhotics who may have ascites. We used the four-variable version of the MDRD formula, which excludes albumin, thus obviating the bias of the therapeutic use of albumin in decompensated cirrhosis. It has been shown that both MDRD formulae have similar precision in patients with chronic kidney disease as well as in cirrhosis (13,15,17).

According to the MDRD formula, there is a correcting factor (0.742) for females, which means that the eGFR in females is lower compared to males with the same Cr concentration, age and race. Thus, the eGFR of a woman is equal to the eGFR of a man with the same Cr, age and race by multiplying 0.742:

image

Then, we considered each female as male and using the MDRD formula for males derived a ‘corrected’ Cr for each female. In other words, we used the MDRD formula two times for each female: first, to calculate the eGFR from the MDRD formula based on the ‘real’ Cr value, and second, we used the MDRD formula in reverse form from the eGFR, in order to calculate the ‘corrected’ Cr, considering each female as male, i.e. using the MDRD formula without the correction factor 0.742. Thus, for each Cr method, females had two Cr values: ‘real’ and ‘corrected.’ MELD scores were separately calculated using each of the four different methods for both ‘real,’ and for females the ‘corrected’ Cr.

We also evaluated the data of 206 consecutive cirrhotic patients on the day of listing for LT, in whom the ‘real’ Cr concentration had been determined using the O'Leary modified Jaffe (mCr) method. As previously described, for each cirrhotic female using the MDRD formula, a ‘corrected’ mCr was calculated. Thus, in this cohort of patients, each female had two mCr values: ‘real’ and ‘corrected.’ Similarly, MELD scores were separately calculated for both ‘real,’ and for females the ‘corrected’ mCr.

Finally, we used the same methodology in 38 consecutive cirrhotic patients assessed for LT. In addition, in this separate cohort, the ‘true’ GFR was estimated using the EDTA-Cr51 method, which is considered an accepted substitute of the ‘gold standard’ inulin clearance (13).

Serum bilirubin was measured using a diazotized sulphanilic method, with caffeine accelerator on the Roche Modular P unit. INR was measured using the ACL FUTURA Coagulometer (Instrumentation Laboratory Ltd, UK), and rabbit brain thromboplastin (PT-Fib HS PLUS, Instrumentation Laboratory), with an international sensitivity index (ISI) of 1.12. For MELD score calculation, the official UNOS web site formula (http://www.unos.org) was used. Three groups were defined using MELD mCr as the reference score: (a) MELD mCr <15 points, (b) between 15 and 24 points and (c) ≥25 points.

Statistical analysis

All data were analyzed using the statistical package SPSS (version 10.0). Quantitative variables were expressed as mean values ± one standard deviation (SD), and/or median values (range). The Mann-Whitney U test was used for comparison of quantitative variables. Significance testing was two sided and set to <0.05. Wilcoxon signed ranks test was used for nonparametric evaluation between paired Cr and GFR values and paired MELD scores. The correlation between GFR derived from different Cr values with MELD scores were evaluated by Spearman correlation.

Results

Patients with abnormal liver function tests

Patient characteristics:  The 158 patients had a median age of 52 (19–92) years, and 94 (58%) were men. The aetiology of liver disease was: alcoholic in 29, viral hepatitis (HBV or HCV) in 43, cryptogenic/NASH in 45, PBC/PSC in 18, autoimmune hepatitis in four, and other causes in 19 patients (Table 2). Sixty-five (41%) were cirrhotics. The 104 patients had one blood sample and 54 had two or more. The 255 samples were obtained from men and 148 from women (Table 1).

Table 2.  Bilirubin, INR, ‘real’ and ‘corrected’ creatinine values, glomerular filtration rate (GFR) and MELD scores between male and female patients with abnormal liver function tests
 Males (n = 94), samples (n = 255) median (range)Females (n = 64), samples (n = 148) median (range)
  1. ‘Corrected’ creatinine was calculated considering each female as male and using the MDRD formula for males (see text).

  2. p > 0.05, *p < 0.002, **p < 0.001 (the findings were similar and for the other three methods of Cr).

  3. GFR = glomerular filtration rate (estimated using the MDRD formula); Cr = serum creatinine; CI = confidence interval; sCr = standard kinetic Jaffe; mCr = O'Leary modified Jaffe; eCr = enzymatic creatinine; cCr = compensated kinetic Jaffe.

Bilirubin (μmol/L)129 (4–913)126 (4-751)
INR1.4 (0.9–8)1.6 (0.9–4.7)
‘Real’ mCr (μmol/L)*[median (range)/mean (95% CI)]”104 (59–1280)/130 (16–145)91 (56–346)/103 (95–110)
‘Corrected’ mCr(μmol/L)*[median (range)/mean (95%CI)]NA115 (73–451)/133 (124–142)
GFR (mCr) (ml/min/1.73 m2) **
[median (range)/mean (95% CI)]68 (4–135)/68 (65–71)58 (12–218)/61 (57–64)
MELD
 sCr20(5–40)19 (5–40)
 mCr,22(5–40)19(5–40)
 eCr20(5–40)19(5–40)
 cCr20(5–40)19(5–40)
Table 1.  Etiology of liver disease in Royal Free Hospital cohort of outpatients or patients admitted to the ward for complications of their liver disease
Cause of liver disease, n (%)Patients n = 158Samples, n = 403
Alcohol29(18.5)129(32)
Viral (hepatitis B or C)43(27)69(17.5)
Autoimmune4(2.5)6(1.5)
Cryptogenic/NASH45(28)89(22)
PBC/PSC18(12)50(12)
Others19(12)60(15)

Laboratory measurements:  The median values of Bil and INR (403 blood samples) were 129 (4–913) μmol/L and 1.5 (0.9–8), respectively, without significant differences between males/females. Median Cr for each of the four different methods in females were significantly lower than males [e.g. mCr (μmol/L): 91(56–346) vs. 104 (59–1280)] (Mann-Whitney test, p < 0.002) (Table 2). Despite this, the median GFR (mL/min/1.73 m2), using the MDRD formula, derived from the four Cr methods (sCr, mCr, eCr and cCr), was also significantly lower in females compared to males [e.g. for GFR-mCr: 58 (12–218) vs. 68 (4–135)] (Mann-Whitney test, p < 0.001) (Table 2).

Correlation between GFR and MELD scores using different methods of Cr measurement:  MELD using mCr had the best negative correlation with its paired GFR using mCr (r = -0.726, p < 0.001) in both males and females, compared to the other three methods (sCr, eCr and cCr), in which the correlation between GFR and MELD was very weak (r = -0.38, -0.18 and -0.25, respectively; p < 0.001).

MELD scores in males and females calculated using different Cr laboratory methods:  Median MELD score was lower in females compared to males for each of the four different methods of Cr measurement: MELD sCr (19 vs. 20), MELD mCr (19 vs. 22), MELD eCr (19 vs. 20) and MELD cCr (19 vs. 20), but none of these differences were significant (p > 0.05) (Table 2).

MELD scores in females calculated using ‘real’ and ‘corrected’ Cr:  In females, the median for each ‘real’ MELD score was significantly lower than the corresponding ‘corrected’ Cr MELD score for each of the four different methods of Cr measurements (Wilcoxon test: all p < 0.001).

Using MELD mCr as the reference score, the median scores of ‘real’ MELD versus ‘corrected’ MELD for each of the four methods of Cr were as follows: (a) MELD 0–14 points (n = 43): all medians were 11 points, (b) MELD 15–24 points (n = 56): 18 versus 19.5 (sCr), 19 versus 21 (mCr), 18 versus 19 (eCr) and 18 versus 19 (cCr), and (c) MELD ≥25 points (n= 49): 27 versus 30 (sCr), 29 versus 31 (mCr), 26 versus 28 (eCr) and 27 versus 28 (cCr) (Table 3).

Table 3.  Median values of ‘real’ and ‘corrected’ Cr MELD scores for different severity of liver disease (mCr is reference creatinine)* in female patients with abnormal liver function tests
 MELD sCr (real/corrected)MELD mCr (real/corrected)MELD eCr (real/corrected)MELD cCr (real/corrected)
  1. ‘Corrected’ creatinine was calculated considering each female as male and using the MDRD formula for males (see text).

  2. sCr = standard kinetic Jaffe; mCr = O'Leary modified Jaffe; eCr = enzymatic creatinine; cCr = compensated kinetic Jaffe.

  3. *The differences between ‘real’ and ‘corrected’ Cr MELD score for each of the four different methods of Cr (sCr, mCr, eCr and cCr) were all significant for all comparisons (p < 0.001).

  4. The proportion of ‘real’ and ‘corrected’ Cr MELD scores, which were different ≥2 and ≥3 points, respectively, progressively increased with higher MELD score:

  5. Ranged between 11–39% and 2–14%, ranged between 18–57% and 11–14%, ranged between 41–98% and 14–51%.

MELD score (using mCr)
 0-1411/1111/1111/1111/11
 15-2418/19.519/2118/1918/19
 ≥2527/3029/3126/2827/28

A difference of ≥2 and ≥3 points, respectively, was found between ‘real’ and ‘corrected’ Cr MELD scores in: (a) 75 (51%) and 37 (25%) [sCr], (b) 93 (67%) and 34 (23%) [mCr], (c) 37 (25%) and 14 (11%) [eCr] and (d) 45 (31%) and 20 (14%) [cCr]. The proportion of ‘real’ and ‘corrected’ Cr MELD scores, which had a difference of ≥2 and ≥3 points, respectively, progressively increased with higher MELD score for all of the four methods of Cr measurement: (a) MELD mCr 0–14 points; range between 11–39% and 2–14%, (b) MELD mCr 15–24 points; range between 18–57% and 11–14% and (c) MELD mCr ≥25; range between 41–98% and 14–51%.

Cirrhotic patients on the waiting list for LT

Patient characteristics:  There were 76 women among the 208 cirrhotics whose median age was 53 (21–73) years; 176 patients were Caucasians and 28 Asians; the aetiology of cirrhosis was: alcoholic in 36, viral hepatitis (HBV or HCV) in 84, cryptogenic/NASH in 19, PBC/PSC in 39, autoimmune hepatitis in eight and other causes in 22 patients (Table 4); 131 (63%) had no or mild ascites and 198 (96%) had no or mild encephalopathy, with no significant differences between males and females.

Table 4.  Etiology of liver disease in Royal Free Hospital cohort of cirrhotic patients listed for liver transplantation
Cause of liver diseasePatients, n (%)
Alcohol36(17.5)
Viral (hepatitis B or C)84(40.5)
Autoimmune8(4)
Cryptogenic/NASH19(9)
PBC/PSC39(18)
Others22(11)

Laboratory measurements:  The median values of urea, sodium, Bil and INR were 4.5 (2.5–22) mmol/L, 136 (120–154) mmol/L, 44 (5–748) μmol/L and 1.5 (0.8–4.9), respectively. Median mCr (μmol/L) in females was 83 (53–241), significantly lower than males [91 (51–335), p = 0.018]. Despite this, the median GFR (ml/min/1.73 m2), using the MDRD formula, was also significantly lower in females compared to males [57 (9–138) vs. 85 (19–191), p < 0.001] (Table 5).

Table 5.  Bilirubin, INR, ‘real’ and ‘corrected’ creatinine values, glomerular filtration rate (GFR) and real and corrected MELD scores (in females with equivalent GFR to males) in a consecutive cohort of male and female cirrhotic patients at listing for liver transplantation at the Royal Free Hospital
 Males (n = 132, samples (n = 132) median (range)Females (n = 76, samples (n = 76) median (range)
  1. ‘Corrected’ creatinine was calculated considering each female as male and using the MDRD formula for males (see text).

  2. *p = 0.018, **p < 0.001.

  3. NA = not applicable; GFR = glomerular filtration rate (estimated using the MDRD formula); mCr = O'Leary modified Jaffe creatinine.

Bilirubin (μmol/L)43 (5–508)44 (6–748)
INR1.5 (0.9–4.2)1.3 (0.8–4.9)
‘Real’ mCr (μmol/L)*[median (range)/mean (95%CI)]91 (51–335)/99 (93–106)83 (53-241)/91 (84-98)
‘Corrected’ mCr (μmol/L) [median (range)/mean (95%CI)]NA109 (69–312)/118 (110–128)
GFR (mCr) (ml/min/1.73 m2) **
[median (range)/mean (95%CI)]85 (19–191)/86 (80–92)57 (9–138)/63 (56–70)
‘real’ MELD (mCr)‘real’/‘corrected’‘real’/‘corrected’
0–1411.7 (7–14)/NA10 (6–14)/ 12 (6–16)
15–2417.7 (15–24)/NA17 (15–23)/ 18.5 (16–25)
≥2531.4 (26–36)/NA28.6 (25–35)/31.1 (28–38)

MELD scores in males and females cirrhotics on the list for LT and MELD scores in females calculated using ‘real’ and ‘corrected’ mCr (Table 5):  Median MELD mCr score was not significantly lower in female candidates for LT compared to male ones (15 vs. 16, p = 0.125). In females, the median ‘real’ MELD mCr score was significantly lower than the corresponding ‘corrected’ MELD mCr score (Wilcoxon test: p < 0.001 for all three groups of ‘real’ MELD mCr score). Between ‘real’ and ‘corrected’ MELD mCr scores a difference of 0, 1, 2 and 3 points was found in 23%, 12%, 36% and 29% of female patients, respectively. There was an increasing difference with increasing MELD scores, but two- to three-point differences also occurred with low MELD scores. Thus, in females, a three-point difference was seen with real MELD scores between 9 and 35 (median 17), a two-point difference with a range 7– 28 (median 14), a one-point difference with a range 7–28 (median 14) and no-point difference with a range 6–18 (median 14). There was no difference greater than three points. The proportion of female cirrhotics, who had a difference of two and three points were: (a) for ‘real’ MELD scores 10–19 points, 42% and 26%, respectively, [median difference 2 (0–3) points], and (b) for ‘real’ MELD scores >19 points, 12.5% and 75%, respectively [median difference 3 (1–3) points].

Cirrhotic patients assessed for LT, in whom ‘true’ GFR was estimated using EDTA-Cr51 patient characteristics:  The 38 cirrhotics had a median age of 50 (29–65) years and 14 were women. The proportion of African American, Caucasians and Asians males/females was similar, compared to the previous cohort; the etiology of cirrhosis was: alcoholic in six, viral hepatitis (HBV or HCV) in 17, PBC/PSC in nine and other causes in six patients.

Laboratory measurements:  The median values of serum albumin, Bil and INR were 32.5 (20–42) g/L, 41 (8–711) μmol/L and 1.4 (1–3.6), respectively, which were not significantly different between men and women. The median mCr (μmol/L) in females was lower, compared to males [65 (39–194) vs. 74 (53–414). Despite this, the median eGFR using the MDRD formula and ‘true’ GFR using EDTA-Cr51, was lower in females (worse renal function) compared to males [MDRD: 90 (25–152) vs. 106 (15–152), and EDTA-Cr51: 81 (16–112) vs. 88 (36–156)] (Table 6). Thus, in our study, although MDRD eGFR overestimated ‘true’ GFR, which is in agreement with previous studies (16,18,19), a lower GFR was still present in females compared to males as measured by EDTA-Cr51 despite a lower Cr. In this small group, the four-variable (4-v) and the original six variable (6-v) versions of the MDRD formula had similar correlation with the ‘true’ GFR (r: 0.64 vs. 0.66, respectively), which was better compared to the correlation between Cockcroft-Gault and ‘true’ GFR (r: 0.55) but again (15) lower, compared to noncirrhotic populations.

Table 6.  Bilirubin, INR, ‘real’ and ‘corrected’ creatinine values, glomerular filtration rate (GFR) and real and corrected MELD scores (in females with equivalent GFR to males) in a separate cohort of 38 cirrhotic patients assessed for liver transplantation at the Royal Free Hospital, in whom ‘true’ GFR was evaluated using EDTA-Cr51
 Males (n = 24)Females (n = 14)
  1. ‘Corrected’ creatinine was calculated considering each female as male and using the MDRD formula for males (see text).

  2. *p = 0.39, **p = 0.54, ***p = 0.11.

  3. NA = not applicable; GFR = glomerular filtration rate (estimated using the MDRD formula and EDTA-Cr51); mCr = O'Leary modified Jaffe creatinine.

Bilirubin (μmol/L), median (range)38 (10–711)57 (8–540)
INR, median (range)1.4 (1–3.6)1.4 (1–2.3)
GFR (mCr)
[median (range)/mean (95%CI)106 (15–152)/98 (85–111)90 (25–152)/90 (69–111)
MDRD (ml/min/1.73 m2)*88 (36–156)/84 (69–96)81 (16–112)/72 (57–91)
EDTA-Cr51 **
‘Real’ mCr (μmol/L)
[Median (range)/mean (95%CI)]***74 (53–414) / 95 (63–126)65 (39–194) / 76 (53–99)
‘Corrected’ mCr (μmol/L)NA87.7 (51–251)

Discussion

The MELD score is a prognostic system, based on three widely available and reproducible variables: Bil, Cr and INR (2,20). The initial model was based on 231 cirrhotics, who had undergone transjugular intrahepatic portosystemic shunt (TIPS) (21). Its prognostic accuracy for short-term mortality in cirrhotics has been confirmed by several studies (3). MELD score has been adopted to prioritize candidates for LT in the United States. The first results of the new MELD-based allocation system were encouraging and considered a step in the right direction (22), ensuring a ‘justice system’ was functioning, allowing equality of access to LT.

However, MELD score is not without limitations. Although the MELD score is considered an evidence-based system based on objective variables, it has been shown that different laboratory assays for INR and Cr give rise to significantly different MELD scores, which potentially lead to inequalities in prioritization of candidates, especially those with the highest priority for LT (5,8).

In addition, although renal dysfunction is associated with increased mortality in cirrhotics (7,23) and the superiority of MELD, which uses Cr, may relate to the impact of renal dysfunction on mortality (24); Cr provides only a rough estimation of GFR and renal function (9,11). In clinical practice, Cr concentration is often discrepant in relation to the ‘true’ GFR. This is due to the influence of several factors unrelated to renal function on Cr concentration, such as dietary intake of creatine (meat ingestion), tubular secretion of Cr from the kidneys and the total pool of body creatine (9). The latter is a well-known extrarenal factor associated with discrepancies in Cr concentration between individuals who have the same renal function (same GFR) but who have different total muscle mass, which can vary according to age, gender, race and nutritional status. The different formulae using Cr for GFR estimation, such as the MDRD, have tried to overcome the influence of these extrarenal factors, including gender on GFR estimation. Although these formulae have several limitations (13,14), which also pertain to cirrhotic patients (15), and in addition, they have been derived in patients with chronic renal failure and not in cirrhotics with renal dysfunction, they provide a more accurate evaluation of the true GFR compared to Cr alone. In addition, MDRD is considered the best formula for eGFR compared to other Cr-based equations in cirrhotic patients (15).

Although bilirubin concentrations are higher in men, compared to women in the general US population (25), in our study, Bil and INR concentrations and age and race were similar between males and females. Thus, the lower MELD scores in females were due to differences in Cr, which were significantly lower than in males. However, females had worse renal function, i.e. significantly lower GFR calculated by the MDRD formula, which takes into account gender, compared to males. Thus, our results, in a cohort of patients with a wide spectrum of liver diseases (from those with only slightly abnormal liver function tests to decompensated cirrhosis and/or candidates for LT), are in accordance with the findings of previous studies in noncirrhotic patients: Cr is not an accurate surrogate marker of renal function and differences remain between males and females (26,27).

In our study we did not evaluate the predictive accuracy of MELD score, but we tried to elucidate the influence of gender in MELD score calculation. Thus, females, despite their significantly worse renal function (lower GFR), had similar MELD scores, compared to males. Although we do not suggest that eGFR would necessarily improve the ability of MELD to predict mortality on the waiting list, we also found that if individual females were considered as males and had their Cr concentrations corrected using the MDRD formula for males, they had higher Cr concentrations for the same GFR and they had significantly higher MELD scores (e.g. 29% of female candidates had a difference of three points between ‘real’ and ‘corrected’ MELD scores). This effect was present irrespective of the Cr method used for measurement. The differences between males and females were confirmed in the small cohort of cirrhotics in whom the ‘true’ GFR was assessed using the EDTA-Cr51 method, i.e. despite a lower median Cr, there was a lower median GFR as measured by EDTA-Cr51 in females, compared to males. In this cohort, the proportions of Asians and Caucasians in the male and female groups were similar to the cohort in which MDRD was calculated, eliminating a potential bias due to ethnicity in our evaluation.

As the MELD allocation system is considered, a primarily justice system, which allocates patients according to severity of liver disease (sickest goes first) (24,28), it is important to assess if there are any biases, particularly of a systematic nature, which may disadvantage some cirrhotics patients versus others. Although the MELD score incorporates Cr as a surrogate marker of renal function, known to be an important prognostic factor in cirrhotics (3,7,24), the lack of distinction between males and females systematically underestimates residual renal function in females with liver disease, which in turn leads to a systematic disadvantage of women with respect to men in terms of prioritization for LT. Thus, we found that 75% of female candidates for LT with a real MELD score >19 had a difference of three points between real and ‘corrected’ MELD scores and 12.5% a two-point difference. Thus, a correction factor of three points could be proposed in females candidates with MELD score >19, in order to correct for the systematic bias. There were no female patients without differences in points, when a ‘real’ MELD was 19 or above.

To our knowledge, our study outlines the problem for the first time. In the original MELD paper (22), gender was not specifically evaluated with respect to MELD calculation. Interestingly, in the 282 cirrhotics, BMI was found not to affect MELD, but females had a lower Cr than males independent of BMI (22). Our findings may explain the results of a recent study, which revealed that women on MELD-based allocation system for LT in the United States have higher mortality and thus are less likely to undergo LT than men (29). This study included 13,484 males and 7554 females from the UNOS registry in the MELD era showing that in both unadjusted and adjusted survival (using 38 transplant variables), there was 18% and 12% decreased survival, respectively, in females, compared to males. Given that renal function is a significant and independent predictor of mortality in cirrhotic patients, i.e. independent of liver function, the discrepancy between Cr concentration and ‘true’ renal function between males and females, despite an equivalent MELD, could be one explanation for this difference. Moreover, a recent paper (30) using UNOS data has shown significantly more males (p < 0.0001) being transplanted following the introduction of MELD (cohort size 13,765 patients) compared to the period before its introduction (cohort size 11,298 patients). In addition, preoperative renal function has been recognized for many years to affect outcome after LT (31), including post-MELD era (30), so that women could also be disadvantaged with respect to posttransplant survival.

Our specific issue about gender differences in Cr and the different eGFR, which are derived from these, has a basic premise with respect to MELD scoring. This is that Cr is completely or almost completely a surrogate of renal function, and does not have specific attributes related to hepatic prognosis other than its renal characteristics. However, even if this were not true, it is unlikely that biologically a ‘nonrenal’ prognostic effect of Cr would be acting differently in male and females. In addition, we assume that mortality on the waiting list, with the same MELD score, is the same for female and male cirrhotic patients (particularly if bilirubin and INR are similar as in our cohorts). Although this has not been documented, it is clear that if it was not true, then this would represent in itself an intrinsic bias in the current MELD system.

Lastly, we cannot prove, by our data, but only suggest, that the issue of gender and Cr may influence mortality on the waiting list and chances of receiving a liver transplant. Thus, we cannot conclusively answer if these differences are clinically relevant, because in our own cohort, the numbers of females with a high-MELD score were few (e.g. ≥20 points were 47), and our waiting list mortality was low (5%). However, two independent publications from the UNOS database can be interpreted as pointing in this direction (29,30), and a future assessment of eGFR (or another measure of GFR), particularly in women with high-MELD scores, will be needed.

We believe that a specific evaluation of male/female differences in creatinine and renal function should take place in a large cohort. If these are confirmed preferably using a ‘true’ and not eGFR measurements, a correction factor for gender should be developed, in order to overcome these inequalities in the current MELD-based allocation system. Alternatively, a more accurate serum marker of renal function instead of creatinine could be used, such as cystatin C. Although cystatin C is not without drawbacks, it has been found to be superior to Cr in a recent metaanalysis (32), to detect earlier renal dysfunction and to reflect more accurately the GFR, independent of diet, sex, age, muscle mass and serum bilirubin, compared to Cr (33,34). The use of a corrected creatinine, or may be true GFR in a defined cohort, should also be explored.

In conclusion, we found that females with liver disease have lower GFR than males for the same Cr in cohorts with abnormal liver function tests as well as candidates for LT. Correcting the Cr in females for the same GFR as in males shows that the current MELD scoring gives rise to significantly lower MELD scores in females despite similar renal function (GFR), and thus a lower priority for LT compared to males, leading to a systematic bias.

Ancillary