World J Surg (2009) 33:1711–1719
DOI 10.1007/s00268-009-0093-4
Value of MELD and MELD-Based Indices in Surgical Risk
Evaluation of Cirrhotic Patients: Retrospective Analysis of 190
Cases
Beatriz P. Costa Æ F. Castro Sousa Æ
Marco Serôdio Æ César Carvalho
Published online: 10 June 2009
Ó Société Internationale de Chirurgie 2009
Abstract
Background Recent studies have suggested that the
Model for End-Stage Liver Disease (MELD) may represent
a promising alternative to the Child-Turcotte-Pugh classification as a predictive factor of operative mortality and
morbidity. This study was designed to evaluate the value
of MELD and four MELD-based indices (iMELD: integrated MELD; MESO: MELD to sodium ratio; MELD-Na:
MELD with incorporation of sodium; MELD-XI: MELD
excluding the International Normalized Ratio) in the
quantification of surgical risk for patients with cirrhosis
and compare its prognostic value with the Child-TurcottePugh classification and two derived scores (proposed by
Huo and Giannini, respectively).
Methods A retrospective study of 190 patients with cirrhosis, operated on in our department between 1993 and
2008, was undertaken.
Results Forty-three percent of patients were included in
Child-Turcotte-Pugh A class, and their mean MELD score
was 12.2 ± 4.9 (range, 6.4–35.2). Mortality and morbidity
rates were 13% and 24%, respectively. In global analysis of
mortality, MELD-based indices presented an acceptable
prognostic performance (auROC = 71–77%), similar to the
three analyzed Child-Turcotte-Pugh-derived scores.
iMELD demonstrated the highest prognostic capacity (auROC = 77%; 95% confidence interval (CI), 66–88; p =
0.0001); operative death probability was 4% (95% CI, 3.6–
4.4) when the score was inferior to 35, 16.1% (95% CI,
14.4–17.9) between 35 and 45, and 50.1% (95% CI,
B. P. Costa (&) F. Castro Sousa M. Serôdio C. Carvalho
IIIrd Surgical Department, Coimbra University Medical School
and Hospitals, Praceta Prof. Mota Pinto, 3000-075 Coimbra,
Portugal
e-mail: beatrizpcosta@huc.min-saude.pt
42.2–58.1) when superior to 45. In elective surgical procedures, iMELD represented a useful prognostic factor of
operative mortality (auROC = 80%; 95% CI, 63–97; p =
0.044) with significant correlation and better accuracy then
MELD and Child-Turcotte-Pugh-derived indices.
Conclusions In this study, iMELD was a useful predictive parameter of operative mortality for patients with
cirrhosis submitted to elective procedures. Further studies
are necessary to define the relevance of MELD-based
indices in the individual surgical risk evaluation.
Abbreviations
iMELD
Integrated MELD
MESO
MELD to sodium ratio
MELD-Na MELD with incorporation of sodium
MELD-XI MELD excluding the international
normalized ratio
r
Spearman’s correlation coefficient.
Introduction
Hepatic cirrhosis represents a recognized surgical risk
factor that is important to quantify during the preoperative
evaluation. In 2003, del Olmo et al. showed that in extrahepatic operations, cirrhosis was associated with a higher
operative mortality (16.3 vs 3.5%) [1]. Child-TurcottePugh classification [2, 3], based on values of bilirubinemia,
albuminemia, and prothrombin time, presence, and severity
of ascites and hepatic encephalopathy, is currently used for
surgical risk evaluation. In 1997, Mansour et al. [4] demonstrated, in abdominal surgery, an operative mortality of
10%, 30%, and 82% for A, B, and C classes, respectively
[5]. Nevertheless, this classification, described for more
123
1712
than 30 years, includes five variables empirically selected,
with equal relative value and some subjectivity, leading to
stratification in three heterogeneous categories with low
discriminative capacity [6]. Recent studies have suggested
that the Model for End-Stage Liver Disease (MELD)—a
score calculated from three laboratory variables (serum
bilirubin, creatinine levels, and International Normalized
Ratio (INR)) and initially developed for evaluation of
short-term prognosis of patients submitted to transjugular
intrahepatic portosystemic shunt and used by United Network for Organ Sharing (UNOS) since 2002 in the definition of priorities for liver transplantation—may represent
a functional hepatic reserve parameter and a predictive
factor of operative mortality and morbidity, with good
correlation and, eventually, with a higher prognostic
accuracy than the traditional Child-Turcotte-Pugh classification [1, 3, 7–17]. More recently, other MELD-based
indices have been proposed, including the ‘‘Integrated
MELD’’ (iMELD), the ‘‘MELD with incorporation of
sodium’’ (MELD-Na), and the ‘‘MELD to sodium ratio’’
(MESO), incorporating other independent predictive factors, such as natremia and age, to improve its prognostic
potential [18–21]. Simultaneously, several authors have
suggested modifications of the Child-Turcotte-Pugh classification, including subdivision of classes (Heuman et al.
2005), introduction of a new D class (Huo et al., 2006), and
addition of creatininemia as a sixth variable (Giannini et al.
2004) [22–26].
A retrospective study of 190 patients, who were operated on in our department between 1993 and 2008, was
performed to evaluate the relevance of MELD and four
MELD-based indices (iMELD, MELD-Na, MESO, and
MELD-XI) in the quantification of surgical risk of patients
with cirrhosis and to compare its prognostic value with
Child-Turcotte-Pugh classification and two derived scores.
Materials and methods
Adult patients with cirrhosis operated on at the IIIrd Surgical Department of Coimbra University Hospitals,
between 1993 and 2008, were identified through the
informatics’ medical registry based on ICD-9 (‘‘International Classification of Diseases, 9th Revision Code).
Diagnosis of cirrhosis was confirmed by clinical criteria
(typical signs and symptoms, previous episodes of hepatic
encephalopathy or variceal bleeding), imaging (dysmorphic or atrophic liver, portal hypertension), endoscopic
(esophageal varices), peroperative, and/or histological
suggestive findings. Chronic renal failure under dialyze
therapy and oral anticoagulants were considered exclusion
criteria, such as insufficient data available to determine
MELD and Child-Turcotte-Pugh scores. The study was
123
World J Surg (2009) 33:1711–1719
based on review of clinical charts and results of laboratory
variables at the moment of hospital admission. The variables studied included age, sex, associated diseases, etiology of liver disease, presence and grade of ascites and
hepatic encephalopathy (definition based on West Haven
criteria [27]), and laboratory values (bilirubinemia, albuminemia, creatininemia, natremia, prothrombinemia, and
INR). The type of surgical procedure (urgent vs elective
surgery, abdominal vs extraabdominal surgery, major vs
minor surgery) and the type of anesthesia (general anesthesia vs other types) were recorded. The American Society
of Anesthesiology (ASA) classification, the Charlson’s
age–comorbidity index, and the Surgical Risk Scale score
were determined [28–32]. The first has been accepted, for
study purposes, as a method to describe the risk of death in
perioperative context and is calculated by adding the points
corresponding to several associated diseases and one point
for each decade of age older than 49 years. The Surgical
Risk Scale, described in 2002 by Sutton et al. [31], has
been used as a risk scoring system, using a scale of 3 to 14
points, and incorporates the classifications of the ‘‘Confidential Enquiry Into Perioperative Deaths’’ (elective,
scheduled, urgent, and emergent), the ‘‘British United
Provident Association’’ (minor, intermediate, major, major
plus, and major complex) and ASA for each surgical procedure. Child-Turcotte-Pugh [2, 3] classification was
determined and Child-Turcotte-Pugh-modified scoring
systems, described respectively by Huo et al. [23] and
Giannini et al. [24], were calculated. MELD and four
MELD-based indices were determined: iMELD, MELDNa, MESO index, and MELD-XI (Table 1) [3, 7, 18–21,
33]. The operative mortality and morbidity rates (primary
end points) and duration of postoperative hospitalization
were determined. Perioperative mortality was defined as
death within 30 days or during the hospitalization after the
operation. Complications were considered to be any deviation from the normal postoperative course other than
normal sequelae (inherent to a specific surgical procedure),
failure of therapy (goal of treatment not attained), or death.
Data management and statistical analysis were performed with SPSSÒ version 15 for Windows (SPSS, Inc.,
Chicago; IL), including v2 and Student’s t tests, logistic
regressions, Spearman’s correlations, ROC (receiver
operating characteristic) curves, and Analyse It (comparison of ROC curves). For all tests, P value \ 0.05 was
considered statistically significant. The study was conformed to the standards of the Declaration of Helsinki.
Results
The study included 190 patients: 81% were male and the
mean age was 61 years (Table 2). The etiology of cirrhosis
World J Surg (2009) 33:1711–1719
1713
Table 1 Formulas of the analyzed MELD-based and Child-Turcotte-Pugh-modified scores
Score
Study
Formula
MELD
Freeman [3]
9.57 Loge creatininemia mg/dl ? 3.78 Loge bilirubinemia mg/dl ? 11.2 Loge ‘‘International
Normalized Ratio’’ ? 6.43
iMELD
Luca et al. [18]
MELD? 0.3 age years - 0.7 natremia mEq/L?100
Minimal values were set to 1.0 and maximal serum creatinine level was considered 4.0 mg/dl
MELD-Na
Biggins et al. [19]
MELD? 1.59 (135-natremia mEq/L)
MESO
Huo et al. [20]
(MELD/natremia mEq/L)100
Maximum and minimum natremia values of 135 and 120 mEq/L, respectively
MELD-XI
Modified Child’s
score
Heuman et al. [33]
Huo et al. [23]
5.11 Loge bilirubinemia mg/dl ? 11.76 Loge creatininemia mg/dl ? 9.44
Included an additional class D (16–18 points) resulting from attribution of one point when
albuminemia \2.3 g/dl, bilirubinemia [8 mg/dl, or prothrombin time prolongation [11 s
Creatinine-modified
Child’s score
Giannini et al. [24]
Estimated by adding one to three points to the original score: 1 if creatininemia B media of the
cohort; 2 if between median and median ? standard deviation; 3 if C to median ? standard
deviation
Table 2 Characteristics of patients with cirrhosis who underwent
operation (n = 190)
n
(%)
Male gender
153
81
Age (yr)
61.4 ± 12 (32–94)
Etiology of cirrhosis
Alcoholic
165
87
Viral
12
6
Other
13
7
ASA classificationm
II
46
24
III
91
48
IV
53
28
Charlson’s modified indexa
6.5 ± 2.5 (3–15)
Data are (%) or median ± standard deviation
a
Charlson’s modified index: summation of points from comorbid
conditions: (1) myocardial infarction, congestive heart failure,
peripheral vascular disease, cerebrovascular disease, dementia,
chronic pulmonary disease, ‘‘connective tissue disease’’, peptic ulcer,
uncomplicated diabetes mellitus; (2) moderate-severe renal disease,
diabetes mellitus with end organ damage, hemiplegia, neoplasia,
leukemia, lymphoma; (3) moderate-severe liver disease; (6) metastatic solid tumor, AIDS; one point for each decade of age older than
49 years (Charlson et al. [30])
was alcoholic in 87% of the cases. Forty-eight percent of
patients were included in ASA III class, and the mean value
of Charlson’s age–comorbidity index was 6.5. Forty-three
percent of patients were included in Child-Turcotte-Pugh’s
A class, and the mean value of MELD was 12 (Table 3).
Hepatocellular carcinomas, abdominal wall hernias, colorectal carcinomas, and gallstone disease represented the
main elective surgical indications. Strangulated hernias,
complicated peptic ulcers, acute cholecystitis, and complicated colorectal carcinomas predominated in urgent cases
(Table 4). Surgical treatment was elective in 60% of cases
and ‘‘major’’ in 77%; it included digestive operations in
73% and was performed under general anesthesia in 98%.
Surgical procedures included hernioplasties (26%), hepatic
resections (17%), cholecystectomies (12%), and colorectal
resections (12%). The mean Surgical Risk Scale score was
8.9 (Table 5). Operative mortality and morbidity rates were
13% and 24%, respectively; mean postoperative hospitalization was 12.9 days (Table 6).
For univariate analysis, the mortality rate was related
with statistical significance to the Child-Turcotte-Pugh
classification, modified Child-Turcotte-Pugh score (Huo
et al. 2006 [23]), creatinine-modified Child-Turcotte-Pugh
score (Giannini et al. 2004 [24]), MELD-derived indices
and to patient’s age, ASA classification, and urgent operations (Table 7). In fact, the mean MELD score was higher
in cases of operative death (17.3 ± 7 vs 11.5 ± 4; p =
0.0001). Patients with MELD C 15 (26%) had higher
mortality rate (31.3 vs 6.1%; p = 0.0001; odds ratio = 5.1;
accuracy = 77.1%). Similarly, mean Child-Turcotte-Pugh’s
score was higher in cases of operative death (9.2 ± 2.5 vs
7.4 ± 2; p = 0.0001). Child-Turcotte-Pugh’s class C
patients (19%) had higher mortality rate (31.4 vs 9.2%; p =
0.0001; odds ratio = 3.4; accuracy = 79.8%). Morbidity
was not significantly related to any studied variables except
age (Table 7). For multivariate analysis, age older than
70 years (p = 0.014; Exp(B) = 5.4; 95% confidence interval
(CI), 1.4–20.8) and albuminemia B 3.1 g/dl (p = 0.049;
Exp(B) = 3.6; 95% CI, 1–13.2) were independent factors of
mortality. Globally, in the study of ROC curves for
mortality, MELD-based indices presented an acceptable
prognostic performance (area under the receiver operating
characteristic curve (auROC) = 71–77%), similar to the
three analyzed Child-Turcotte-Pugh-derived scores (auROC = 72–75%); only the difference between MESO and
MELD-Na was statistically significant (p = 0.0433;
Table 8; Fig. 1). iMELD demonstrated the highest
123
1714
Table 3 Child-Turcotte-Pugh
classification, Child-TurcottePugh-modified scores, MELD,
and MELD-based indices of
cirrhotic patients submitted to
surgery (n = 190)
World J Surg (2009) 33:1711–1719
Study
n
(%)
A
82
43
B
72
38
C
36
Child-Turcotte-Pugh classification
Modified Child-Turcotte-Pugh score
Data are (%) or median ±
standard deviation
7.6 ± 2.2 (5–14)
Freeman [3]
Huo et al. [23]
19
7.8 ± 2.4 (5–15)
Creatinine-modified Child-Turcotte-Pugh score
Giannini et al. [24]
9 ± 2.5 (6–16)
MELD
Freeman [3]
12.2 ± 4.9 (6.4–35.2)
35.7 ± 7 (21.1–57.3)
iMELD
Luca et al. [18]
MELD-Na
Biggins et al. [19]
14.8 ± 7.4 (6.4–36.6)
MESO Index
Huo et al. [20]
9.1 ± 3.7 (4.2–24)
MELD-XI
Heuman et al. [33]
13.1 ± 4.4 (9.4–28.4)
Table 4 Indications for surgery in 190 patients with cirrhosis
Table 5 Surgical treatment in 190 patients with cirrhosis
n
(%)
Hepatocellular carcinoma
34
18
Abdominal wall hernia
25
Colorectal carcinoma
Gallstone disease
n
(%)
Elective
113
60
‘‘Major’’
146
77
13
Under general anesthesia
187
98
19
16
10
8
Type of procedure
100
53
8
4
Extradigestive
53
27
37
20
Elective
Gastric carcinoma
Digestive extrahepatic
Portal hypertension
3
2
Hepatic
Status after Hartmann operation
2
1
Interventions
Other
6
3
Hernioplasty/herniorrhaphy
56
26
Hepatic resection
36
17
Strangulated hernia of abdominal wall
19
10
Peptic ulcer (perforation/hemorrhage)
19
10
Cholecystectomy
Colo/rectal resection
27
26
12
12
Acute cholecystitis
9
5
Gastrectomy
15
7
Colorectal carcinoma (obstruction/perforation)
7
4
Explorative laparotomy/laparoscopy
8
4
Esophagus/gastric variceal bleeding
5
3
Troncular vagotomy and pyloroplasty
7
3
Abdominal trauma
5
3
Enterectomy
5
2
13
7
Gastrojejunostomy
5
2
Porto-caval shunt
5
2
28
13
Urgent
Other
prognostic capacity (auROC = 77%, 95% CI, 66–88; p =
0.0001); its mean value was higher in cases of operative
death (42.8 ± 8.5 vs 34.6 ± 6.1; p = 0.0001). Patients with
iMELD score C 40 (28%) had higher mortality rate (29.8
vs 6.6%; p = 0.0001; odds ratio = 4.5; accuracy = 75.6%).
Operative death predicted probability, calculated from the
logistic regression model, increased with iMELD (from
0.8–77.8%): 4% when inferior to 35 (95% CI, 3.6–4.4),
16.1% (95% CI, 14.4–17.9) between 35 and 45, and 50.1%
(95% CI, 42.2–58.1) when [ 45 (Fig. 2). Regarding morbidity, predictive performance of the studied cirrhotic
variables was low (auROC = 41–51%; not significant).
In the analysis of mortality after elective surgical procedures, MELD-derived indices increased its prognostic
advantages and iMELD was revealed to be a good predictive parameter of operative mortality (auROC = 80%,
123
Others
Surgical Risk Scale score (Sutton et al., 2002): 8.9 ± 1.9 (5–13)
95% CI, 63–97; p = 0.044) with better predictive potential
than MELD (p = 0.0226) and the three studied ChildTurcotte-Pugh-derived indices (p = 0.0005, p = 0.0003, and
p = 0.001 compared with traditional, Huo, and Giannini
scores, respectively); relations between all other variables
did not reach statistical significance except MELD XI vs
Child-Turcotte-Pugh-derived indices (Table 9). The
iMELD score was superior in cases of mortality (40.1 ±
7.3 vs 33 ± 5.6; p = 0.016); in elective procedures, operative death probability was 0.7% (95% CI\ 0.6–0.8) when
the score was \ 30, 3.2% (95% CI, 2.7–3.8) between 30
and 40, and 17.2% (95% CI, 11.1–23.3) when [ 40.
World J Surg (2009) 33:1711–1719
1715
Table 6 Results of surgical treatment in 190 patients with cirrhosis
Mortality
n
(%)
25
13
Sepsis
9
Hepatic insufficiency
8
Disseminated intravascular coagulation
3
Irreversible hypovolemic shock
3
Digestive hemorrhage
1
Acute pulmonary edema
1
Morbidity
46
Operative wound complications
22
Nosocomial pneumonia
6
Organ/space infection
Pleural effusion
5
3
Intraperitoneal hemorrhage
3
Other
7
Postoperative hospitalization (days)
24
12.9 ± 13.4 (1–92)
Data are (%) or median ± standard deviation
Furthermore, the correlations between MELD and
iMELD compared with Child-Turcotte-Pugh score were
significant (Spearman’s r = 0.72; p = 0.0001 and Spearman’s r = 0.58; p = 0.0001, respectively). The correlation
between MELD-XI and MELD was very good (Spearman’s r = 0.88; p = 0.0001).
Discussion
The particular relevance of our study relies on the direct
comparison, in the perioperative context, of five MELDbased indices and three Child-Turcotte-Pugh derived
scores.
In this series, the population was characterized by high
prevalence of associated diseases: the Charlson’s agecomorbidity index was 6.5 ± 2.5 (3–15) and 48% of
patients were included in ASA III class. In relation to
cirrhosis severity, 57% of our patients were included in
Child-Turcotte-Pugh’s B and C classes and 34% had an
MELD score [ 12. Surgical operations were mainly elective (60%) and digestive (73%, including 20% hepatic
procedures) and were associated with a high mean Surgical
Risk Score (8.9). Operative mortality (13%) and morbidity
rates (24%) were similar to those published in recent literature (mortality between 9.8 and 28%) [9, 11].
Globally, this study suggested that MELD and MELDbased indices are significant predictive parameters of
operative mortality in patients with cirrhosis similar to
Child-Turcotte-Pugh classification-derived scores (auROC
= 71–77% and 72–75%, respectively). In fact, iMELD
demonstrated the highest prognostic capacity, although the
differences were not statistically significant.
Interestingly, the prognostic advantages of MELD-based
indices improved when considering only elective operations. In this group, iMELD represented a useful prognostic
factor of operative mortality (auROC = 80%) with better
predictive potential than MELD and Child-Turcotte-Pughderived scores. In reality, the operative risk evaluation
seems to be much more important in the elective context.
Nevertheless, MELD variables are susceptible to potential
interferences that are associated with acute diseases (e.g.,
sepsis, dehydration, trauma coagulopathy), which do not
reflect the severity of the underlying liver disease.
The potential interest of MELD as a predictive factor of
postoperative outcome for patients with cirrhosis has been
confirmed by several authors in a wide variety of surgical
operations [8, 9, 34]. In a retrospective study of 772
patients submitted to major digestive, orthopedic, and
cardiac surgery, published in 2007 by Teh et al. [8], the
operative mortality ranged from 5.6% when MELD score
was \ 8 to more than 50% when MELD score was [ 20.
There also was a mortality increase of 14% on the first and
third months for each additional point in MELD score, of
15% during the first year and 6% during subsequent years;
concordance (c-statistic) relative to first and third months
were 0.78 and 0.82, respectively. Northup et al. reported, in
a retrospective series of 130 patients undergoing general,
orthopedic, cardiovascular, and urologic procedures, published in 2005, that operative mortality risk increased 1%
for each additional point for MELD scores between 5 and
20 and 2% for values [20 (c = 0.72) [9, 35]. Other studies,
such as Befeler et al. [11], Farnsworth et al. [12], and
Perkins et al. [13], verified an increased risk of mortality
and morbidity after surgery, mainly abdominal, in cases of
MELD[8 to 14. Cucchetti et al. [10] reported that MELD
[ 11 signifies an elevated risk of morbidity after hepatic
resection for hepatocellular carcinoma (auROC = 85%),
particularly hepatic insufficiency compared with values\9
(83.3 vs 8.1%, respectively). On the contrary, some studies
did not notice advantages of MELD compared with ChildTurcotte-Pugh classification, such as Hoteit et al. [36] in a
study of 195 patients submitted to extrahepatic surgery,
Schroeder et al. [37] in elective hepatic operations, and
others [38, 39].
Hyponatremia has been associated with severe complications of cirrhosis, including ascites, hepatorenal syndrome, and liver-related mortality [14]. Several authors
have verified that natremia constituted an independent
prognostic factor of mortality in cirrhotic candidates for
transplantation and proposed its inclusion into MELD,
creating new scores, e.g., iMELD, MELD-Na, and MESO
[18–21, 40]. Luca et al. [18] showed that iMELD was
associated with an increase in the accuracy of prediction of
123
1716
World J Surg (2009) 33:1711–1719
Table 7 Univariate analysis of operative mortality and morbidity of 190 patients with cirrhosis who underwent operation
Mortality
p
Morbidity
p
Sex (male/female)
13.1 vs 13.5%
NS
25.5 vs 18.9%
NS
Age (year)
67.1 ± 11.3 vs 60.6 ± 11.8
0.01
57.7 ± 13.7 vs 62.6 ± 11.1
0.016
Cirrhosis alcoholic/viral/other
13.3 vs 8.3 vs 15.4%
NS
24.2 vs 41.7 vs 7.7%
NS
ASA classification (IV/III/II)
27.8 vs 8.9 vs 4.3%
0.001
18.5 vs 24.4 vs 30.4%
NS
Charlson’s modified index
7.3 ± 2.5 vs 6.4 ± 2.5
NS
6.2 ± 2.3 vs 6.6 ± 2.5
NS
Albuminemia (g/dl)
2.8 ± 0.7 vs 3.6 ± 0.7
0.0001
3.6 ± 0.7 vs 3.4 ± 0.8
NS
INR
1.8 ± 0.9 vs 1.3 ± 0.2
0,0001
1.3 ± 0.2 vs 1.4 ± 0.4
NS
Creatininemia (mg/dl)
1.5 ± 0.7 vs 1 ± 0.4
0.0001
1.1 ± 0.4 vs 1.1 ± 0.5
NS
Bilirubinemia (mg/dl)
2.4 ± 2.6 vs 1.8 ± 2.9
NS
1.8 ± 1.8 vs 1.9 ± 3.1
NS
Natremia (mEq/l)
135.3 ± 7.7 vs 136.2 ± 4.7
NS
136.2 ± 5.6 vs 136 ± 5.1
NS
Ascites (yes/no)
Encephalopathy (yes/no)
26 vs 4.4%
14 vs 14%
NS
NS
19.6 vs 28.1%
22.2 vs 22.4%
NS
NS
Child’s classification (C/B/A)
31.4 vs 14.1 vs 4.9%
0.001
25.7 vs 19.7 vs 26.8%
NS
Child’s score
9.2 ± 2.5 vs 7.4 ± 2.0
0.0001
7.5 ± 2.2 vs 7.7 ± 2.2
NS
Modified Child’s scorea
9.5 ± 2.8 vs 7.5 ± 2.2
0.0001
7.6 ± 2.4 vs 7.8 ± 2.4
NS
Creatinine-modified Child’s scoreb
11.2 ± 3 vs 8.7 ± 2.2
0.0001
8.8 ± 2.4 vs 9.1 ± 2.5
NS
MELD
17.3 ± 7 vs 11.5 ± 4
0.0001
11.9 ± 3.9 vs 12.3 ± 5.2
NS
iMELD
42.8 ± 8.5 vs 34.6 ± 6.1
0.0001
34 ± 6.4 vs 36.2 ± 7.1
NS
MELD-Na
21.8 ± 9.3 vs 13.8 ± 6.5
0.001
14.6 ± 7.7 vs 14.9 ± 7.3
NS
MESO
12.9 ± 5.1 vs 8.6 ± 3.1
0,001
8.9 ± 3 vs 9.2 ± 3.9
NS
MELD-XI
16.7 ± 6 vs 12.5 ± 3.8
0.003
12.9 ± 4.1 vs 13.2 ± 4.4
NS
Surgical Risk Scale
8.4 ± 2.4 vs 9 ± 1.8
NS
9.2 ± 1.8 vs 8.8 ± 2
NS
Urgent procedure (yes/no)
26 vs 4.4 %
0.0001
23.4 vs 24.8 %
NS
‘‘Digestive’’ procedure (yes/no)
15.2 vs 7.7 %
NS
26.1 vs 19.2 %
NS
‘‘Major’’ procedure (yes/no)
15.1 vs 6.8 %
NS
25.3 vs 20.5 %
NS
General anesthesia (yes/no)
13.4 vs 0 %
NS
24.6 vs 0 %
NS
Date (1993-1998/1999-2003/2004-2008)
13 vs 10.9 vs 15.2 %
NS
17.4 vs 29.1 vs 27.3 %
NS
Data are (%) or median ± standard deviation
INR International normalized ratio
a
Huo et al. [23], 2006
b
Giannini et al. [24], 2004
mortality after transjugular intrahepatic portosystemic
shunt (auROC = 78% vs 69%) and on the transplantation
waiting list (auROC = 81% vs 75%), particularly in
patients with MELD \ 15. The incorporation of natremia
as a postoperative prognostic factor may be particularly
important in cases with lower MELD scores (such as in our
series) but may have disadvantages, namely, the susceptibility of natremia values to important oscillations with
simple therapeutic measures (e.g., administration of
diuretics), which do not always reflect modifications of the
severity of liver disease.
In our study, MELD-XI revealed a good correlation
with MELD and similar statistic concordance in overall
and elective surgery cohorts; its performance was significantly better than Child-Turcotte-Pugh-derived scores in
elective procedures. MELD-XI, calculated only with
serum bilirubin and creatinine levels, was originally
123
proposed in 2007 to prioritize cirrhotic candidates for
liver transplantation under therapeutic anticoagulation and
it may be especially useful in cases of possible dissociation of INR in relation to liver function (as occurs in
cholestasis, pancreatic insufficiency, intestinal malabsorptive diseases, and disseminated intravascular coagulation, among others) and to monitor and correct
discrepancies associated with the interlaboratory variability of INR measurements [33, 41].
Finally, there are several limitations of this study: retrospective design, potential selection bias (e.g., unknown
number of patients with indication for surgery who were
not operated on because of high surgical risk), low
dimension of the sample, heterogeneity of the surgical
procedures, and limited number of analyzed variables.
Other aspects not included in the study are the influence of
therapeutic maneuvers on MELD parameters and the
World J Surg (2009) 33:1711–1719
1717
Table 8 Areas under the ROC curves for operative mortality and morbidity in 190 patients with cirrhosis
Mortality vs
Morbidity vs
AUC (%)
95% CI
p
AUC (%)
95% CI
p
Charlson’s modified index
62
50–74
NS
53
42–64
NS
Surgical Risk Scale
42
28–57
NS
55
46–65
NS
Child’s score
72
61–83
0.001
48
37–58
NS
Modified Child’s scorea
73
62–84
0.0001
47
37–58
NS
Creatinine-modified Child’s scoreb
75
63–87
0.0001
47
37–58
NS
MELD
76
64–89
0.0001
51
42–60
NS
iMELD
MELD-Na
77
76
66–88
64–87
0.0001
0.0001
41
48
31–52
38–58
NS
NS
MESO
76
63–88
0.0001
51
41–60
NS
MELD-XI
71
59–83
0.001
51
41–60
NS
AUC area under the ROC curve, CI confidence interval
a
Huo et al. [23], 2006
b
Giannini et al. [24], 2004
Fig. 1 Comparison of ROC curves corresponding to Child-TurcottePugh score (auROC = 72%, p = 0.001), MELD (auROC = 76%; p =
0.0001), and iMELD (auROC = 77%; p = 0.0001) as predictive
factors of operative mortality in 190 patients with cirrhosis
eventual advantage of serial determinations (e.g., Delta
MELD proposed in 2005 by Huo et al. [42]). A recent
report by Lisman et al. [41] demonstrated clinically relevant interlaboratory variations of the calculated MELD,
resulting mainly of differences on INR and creatinine
measurements that justify adjustments and normalization.
Further refinement and validation of MELD-based
indices are needed and must include minimization of
potential interferences into variables, due to different laboratory methods, sex and ethnic variations, acute occurrences (e.g., sepsis and dehydration), and therapeutic
management. Other future investigational perspectives in
Fig. 2 Relationship between the predicted probability of operative
death and iMELD in 190 patients with cirrhosis who underwent
operation
this area will be to improve individual risk evaluation with
definitions of ‘‘cutoffs,’’ correlation with quality of life and
performance in cases of ascites, chronic encephalopathy,
hepatopulmonary
syndrome,
and
portopulmonary
hypertension.
Therefore, well-designed prospective studies, including
a higher number of patients, specific surgical procedures,
and analysis of additional variables are needed to clarify
the interest and applicability of MELD-based indices on
the daily clinical practice, particularly during the preoperative period.
123
1718
World J Surg (2009) 33:1711–1719
Table 9 Areas under the ROC curves for mortality and morbidity in 113 patients with cirrhosis submitted to elective surgery
Mortality vs
Morbidity vs
AUC (%)
95% CI
p
AUC (%)
95% CI
p
Charlson’s modified index
78
62–94
0.034
55
36–56
NS
Surgical risk scale
64
27–99
NS
56
44–67
NS
Child’s score
54
24–84
NS
44
29–59
NS
Modified Child’s score
a
54
24–84
NS
44
29–58
NS
Creatinine-modified Child’s scoreb
53
21–84
NS
44
29–59
NS
MELD
61
27–94
NS
57
45–68
NS
iMELD
MELD-Na
80
62
63–97
28–96
0.044
NS
41
49
28–55
36–61
NS
NS
MESO
53
14–91
NS
54
41–67
NS
MELD-XI
62
29–95
NS
55
44–67
NS
AUC area under the ROC curve, CI confidence interval, NS not significant
a
Huo et al. [23], 2006
b
Giannini et al. [24], 2004
Conclusions
In this series, iMELD was revealed to be a useful predictive parameter of operative mortality in patients with cirrhosis submitted to elective procedures, with significant
correlation with three studied Child-Turcotte-Pugh-derived
scores (traditional, Huo et al., and Giannini et al., respectively) and MELD. Its simplicity, objectivity, and reproducibility may constitute additional advantages in relation
to Child-Turcotte-Pugh-derived scores. Nevertheless, further studies are necessary to demonstrate the eventual
interest of MELD-based indices in the surgical risk evaluation of patients with chronic hepatic diseases.
References
1. del Olmo JA, Flor-Lorente B, Flor-Civera B et al (2003) Risk
factors for nonhepatic surgery in patients with cirrhosis. World J
Surg 27:647–652
2. Pugh RN, Murray-Lyon IM, Dawson JL et al (1973) Transection
of the oesophagus for bleeding oesophageal varices. Br J Surg
60:646–649
3. Freeman RB (2005) MELD: the holy grail of organ allocation? J
Hepatol 42:16–20
4. Mansour A, Watson W, Shayani V et al (1997) Abdominal
operations in patients with cirrhosis: still a major surgical challenge. Surgery 122:730–736
5. O’Leary JG, Friedman LS (2007) Predicting surgical risk in
patients with cirrhosis: from art to science. Gastroenterology
132:1609–1610
6. Durand F, Valla D (2005) Assessment of the prognosis of cirrhosis: Child-Pugh versus MELD. J Hepatol 42:S100–S107
7. Malinchoc M, Kamath PS, Gordon FD et al (2000) A model to
predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts. Hepatology 31:864–871
8. Teh SH, Nagorney DM, Stevens SR et al (2007) Risk factors for
mortality after surgery in patients with cirrhosis. Gastroenterology 132:1261–1269
123
9. Northup PG, Wanomaker RC, Lee VD et al (2005) Model for
End-Stage Liver Disease (MELD) predicts nontransplant surgical
mortality in patients with cirrhosis. Ann Surg 242:244–251
10. Cucchetti A, Ercolani G, Vivarelli M et al (2006) Impact of
model for end-stage liver disease (MELD) score on prognosis
after hepatectomy for hepatocellular carcinoma on cirrhosis.
Liver Transpl 12:966–971
11. Befeler AS, Palmer DE, Hoffman M et al (2005) The safety of
intra-abdominal surgery in patients with cirrhosis: model for endstage liver disease is superior to Child-Turcotte-Pugh classification in predicting outcome. Arch Surg 140:650–654
12. Farnsworth N, Fagan SP, Berger DH et al (2004) Child-TurcottePugh versus MELD score as a predictor of outcome after elective
and emergent surgery in cirrhotic patients. Am J Surg 188:580–583
13. Perkins L, Jeffries M, Patel T (2004) Utility of preoperative
scores for predicting morbidity after cholecystectomy in patients
with cirrhosis. Clin Gastroenterol Hepatol 2:1123–1128
14. Huo TI, Lee SD, Lin HC (2008) Selecting an optimal prognostic
system for liver cirrhosis: the model for end-stage liver disease
and beyond. Liver Int 28:606–613
15. Kamath PS, Kim WR, Advanced liver disease study group (2007)
The model for end-stage liver disease (MELD). Hepatology
45:797–805
16. Durand F, Valla D (2008) Assessment of prognosis of cirrhosis.
Semin Liver Dis 28:110–122
17. Cholongitas E, Papatheodoridis GV, Vangeli M et al (2005)
Systematic review: the model for end-stage liver disease–should
it replace Child-Pugh’s classification for assessing prognosis in
cirrhosis? Aliment Pharmacol Ther 22:1079–1089
18. Luca A, Angermayr B, Bertolini G et al (2007) An integrated
MELD model including serum sodium and age improves the
prediction of early mortality in patients with cirrhosis. Liver
Transp 13:1174–1180
19. Biggins SW, Kim WR, Terrault NA et al (2006) Evidence-based
incorporation of serum sodium concentration into MELD. Gastroenterology 130:1652–1660
20. Huo TI, Wang YW, Yang YY et al (2007) Model for end-stage
liver disease score to serum sodium ratio index as a prognostic
predictor and its correlation with portal pressure in patients with
liver cirrhosis. Liver Int 27:498–506
21. Huo T, Lin H, Huo SC et al (2008) Comparison of four model for
end-stage liver disease-based prognostic systems for cirrhosis.
Liver Transpl 14:837–844
World J Surg (2009) 33:1711–1719
22. Heuman DM, Habib A, Abou-Assi S et al (2005) Rationally
derived Child-Turcotte-Pugh (CTP) subclasses permit accurate
stratification of near-term risk in cirrhotic patients referred for
liver transplantation. Gastroenterology 128:A-734
23. Huo TI, Lin HC, Wu JC et al (2006) Proposal of a modified
Child-Turcotte-Pugh scoring system and comparison with the
model for end-stage liver disease for outcome prediction in
patients with cirrhosis. Liver Transpl 12:65–71
24. Giannini E, Botta F, Fumagalli A et al (2004) Can inclusion of
serum creatinine values improve the Child-Turcotte-Pugh score
and challenge the prognostic yield of the model for end-stage
liver disease score in the short-term prognostic assessment of
cirrhotic patients? Liver Int 24:465–470
25. Papatheodoridis GV, Cholongitas E, Dimitriadou E et al (2005)
MELD vs. Child-Pugh and creatinine-modified Child-Pugh score
for predicting survival in patients with decompensated cirrhosis.
World J Gastroenterol 11:3099–3104
26. Angermayer B, Koening F, Cejna M et al (2002) Creatininemodified Child-Pugh score (CPSC) compared with MELD-score to
predict survival in patients undergoing TIPS. Hepatology 36:860A
27. Ferenci P, Lockwood A, Mullen K et al (2002) Hepatic encephalopathy: definition, nomenclature, diagnosis, and quantification:
final report of the working party at the 11th world congress of
gastroenterology, Vienna, 1998. Hepatology 35:716–721
28. ASA physical status classification system. American Society of
Anesthesiologists website. http://www.asahq.org/clinical/physical
status.htm.2008. Accessed 28 Feb 2008
29. Lorimer JW, Doumit G (2007) Comorbidity is a major determinant of severity in acute diverticulitis. Am J Surg 193:681–685
30. Charlson ME, Szatrowski TP, Peterson J et al (1994) Validation of a
combined comorbidity index. J Clin Epidemiol 47:1245–1251
31. Sutton R, Bann S, Brooks M et al (2002) The surgical risk scale
as an improved tool for risk-adjusted analysis in comparative
surgical audit. Br J Surg 89:763–768
32. Neary WD, Prytherch D, Foy C et al (2007) Comparison of
different methods of risk stratification in urgent and emergency
surgery. Br J Surg 94:1300–1305
1719
33. Heuman DM, Mihas AA, Habib A et al (2007) MELD-XI: a
rational approach to ‘‘sickest first’’ liver transplantation in cirrhotic
patients requiring anticoagulant therapy. Liver Transpl 13:30–37
34. Teh SH, Sheppard BC, Schwartz J et al (2008) Model for endstage liver disease score fails to predict perioperative outcome
after hepatic resection for hepatocellular carcinoma in patients
without cirrhosis. Am J Surg 195:697–701
35. Durand F (2006) Risk scores in cirrhotic patients: from nontransplant surgery to transplantation and back. J Hepatol 44:620–
621
36. Hoteit MA, Ghazale AH, Bain AJ et al (2008) Model for endstage liver disease score versus child score in predicting the
outcome of surgical procedures in patients with cirrhosis. World J
Gastrenterol 14:1774–1780
37. Schroeder RA, Marroquin CE, Bute BP et al (2006) Predictive
indices of morbidity and mortality after liver resection. Ann Surg
243:373–379
38. Bingener J, Cox D, Michalek J et al (2008) Can the MELD score
predict perioperative morbidity for patients with liver cirrhosis
undergoing laparoscopic cholecystectomy? Am Surg 74:156–159
39. Filsoufi F, Salzberg SP, Rahmanian PB et al (2007) Early and late
outcome of cardiac surgery in patients with liver cirrhosis. Liver
Transp 13:990–995
40. Londoño M, Cárdenas A, Guevara M et al (2007) MELD score
and serum sodium in the prediction of survival of patients with
cirrhosis awaiting liver transplantation. Gut 56:1283–1290
41. Lisman T, van Leeuwen Y, Adelmeijer J et al (2008) Interlaboratory variability in assessment of the model of end-stage liver
disease score. Liver Int 28:1344–1351
42. Huo TI, Wu JC, Lin HC et al (2005) Evaluation of the increase in
model for end-stage liver disease (MELD) score overtime as a
prognostic predictor in patients with advanced cirrhosis: risk
factor analysis and comparison with initial MELD and ChildTurcotte-Pugh score. J Hepatol 42:826–832
123