GASTROENTEROLOGY 2007;132:1261–1269
CLINICAL–LIVER, PANCREAS, AND BILIARY
TRACT
Risk Factors for Mortality After Surgery in Patients With Cirrhosis
*Division of Gastroenterologic and General Surgery, ‡Division of Biostatistics, §Department of Anesthesiology, and 储Division of Gastroenterology and Hepatology,
Mayo Clinic, Rochester, Minnesota
See editorial on page 1609; CME quiz on page
1596.
Background & Aims: Current methods of predicting risk of postoperative mortality in patients with
cirrhosis are suboptimal. The utility of the Model
for End-stage Liver Disease (MELD) in predicting
mortality after surgery other than liver transplantation is unknown. The aim of this study was to
determine the risk factors for postoperative mortality in patients with cirrhosis. Methods: Patients
with cirrhosis (N ⴝ 772) who underwent major
digestive (n ⴝ 586), orthopedic (n ⴝ 107), or cardiovascular (n ⴝ 79) surgery were studied. Control
groups of patients with cirrhosis included 303 undergoing minor surgical procedures and 562 ambulatory patients. Univariate and multivariable proportional hazards analyses were used to determine
the relationship between risk factors and mortality.
Results: Patients undergoing major surgery were at
increased risk for mortality up to 90 days postoperatively. By multivariable analysis, only MELD
score, American Society of Anesthesiologists class,
and age predicted mortality at 30 and 90 days, 1
year, and long-term, independently of type or year
of surgery. Emergency surgery was the only independent predictor of duration of hospitalization
postoperatively. Thirty-day mortality ranged from
5.7% (MELD score, <8) to more than 50% (MELD
score, >20). The relationship between MELD score
and mortality persisted throughout the 20-year
postoperative period. Conclusions: MELD score,
age, and American Society of Anesthesiologists
class can quantify the risk of mortality postoperatively in patients with cirrhosis, independently of
the procedure performed. These factors can be used
in determining operative mortality risk and
whether
elective surgical procedures can be delayed until
after liver transplantation.
P
atients with cirrhosis are believed to be at increased
risk of mortality and morbidity after surgery,1–3 but
the duration of risk for increased mortality has not been
demonstrated. In patients with cirrhosis undergoing various surgical procedures, risk factors for mortality have
been shown to include Child–Turcotte–Pugh (CTP) score
of 7 or greater, elevated serum creatinine concentration,
cardiopulmonary comorbidity, and American Society of
Anesthesiologists (ASA) physical status class of IV or V.4
However, the quantitative risk of mortality at specific
times after surgery has not been characterized. Many
patients with cirrhosis are candidates for liver transplantation, but, at present, there is no way to determine
whether elective procedures in these patients should be
delayed until after liver transplantation. The increased
interest in surgery in patients with liver disease relates in
part to the advent of liver transplantation, which offers
the prospect of long-term survival for patients who
would not previously have been considered for major
surgery.
The Model for End-stage Liver Disease (MELD)— derived from the international normalized ratio (INR), serum total bilirubin, serum creatinine concentration, and
etiology of liver disease—was originally developed to assess the short-term prognosis of patients with cirrhosis
receiving a transjugular intrahepatic portosystemic shunt.5
The MELD subsequently was validated as a predictor of
mortality in patients with many types of liver disease,
even if the etiology was not considered.6 Since February
Abbreviations used in this paper: ASA, American Society of Anesthesiologists; CTP, Child–Turcotte–Pugh; INR, international normalized
ratio; IQR, interquartile range; MELD, Model for End-stage Liver Disease.
© 2007 by the AGA Institute
0016-5085/07/$32.00
doi:10.1053/j.gastro.2007.01.040
CLINICAL–LIVER,
PANCREAS, AND
BILIARY TRACT
SWEE H. TEH,* DAVID M. NAGORNEY,* SUSANNA R. STEVENS,‡ KENNETH P. OFFORD,‡ TERRY M. THERNEAU,‡
DAVID J. PLEVAK,§ JAYANT A. TALWALKAR,储 W. RAY KIM,储 and PATRICK S. KAMATH储
1262
TEH ET AL
CLINICAL–LIVER,
PANCREAS, AND
BILIARY TRACT
2002, the MELD system has been used with excellent
results to prioritize allocation of organs for liver transplantation in the United States.7
The utility of the MELD for predicting mortality in
patients with cirrhosis undergoing surgery other than
liver transplantation has not been well studied. We and
others have shown that the MELD score can be used to
determine which patients with cirrhosis are at low risk of
mortality after hepatic resection for hepatocellular carcinoma.8,9 Several other reports have concluded that patients with a higher MELD score are at increased risk of
mortality or morbidity after surgery.10 –13 These studies,
however, addressed only postoperative morbidity and not
mortality,13 were restricted to cardiac surgery alone,10
included a patient cohort with a low mortality rate,11,12
or addressed only 30-day mortality.12 One study indicated that the MELD was not predictive of mortality after
hepatic resection, but the majority of patients in the
cohort may not have had cirrhosis.14 In the present study,
we investigated short-term and long-term mortality risks
in patients with cirrhosis who underwent various surgical
procedures, excluding liver transplantation.
Patients and Methods
Study Population
Using the Surgical Procedure Index database at
the Mayo Clinic in Rochester, Minnesota, we retrospectively searched for the records of patients with cirrhosis
from any cause who underwent major digestive, orthopedic, or cardiac surgical procedures. Patients were stratified into 2 study populations based on decade of treatment (1980 –1990 vs 1994 –2004). A total of 825 patients
were identified. All patients who had laparoscopic cholecystectomy (n ⫽ 46) and those in whom all MELD
parameters were not measured within 24 hours before
surgery (n ⫽ 7) were excluded. The patients who underwent laparoscopic cholecystectomy were excluded because the risk associated with this procedure in patients
with cirrhosis had been addressed previously in much
larger studies.13,15 The final study group included 772
patients: 426 consecutive patients who had surgery in the
decade 1980 through 1990 (“Era A”) and who were part
of an earlier database, portions of which have been previously analyzed;4 and 346 consecutive patients who had
surgery in 1994 through 2004 (“Era B”). To account for
possible differences in patient population and abstraction method, all analyses were stratified by era.
All included patients had cirrhosis documented by
histologic or intraoperative findings, or the diagnosis was
based on the findings of at least 2 of the following:
(1) gastroesophageal varices on endoscopy; (2) serum
albumin ⱕ3.4 g/dL, INR ⱖ1.3, or serum bilirubin ⱖ2
mg/dL; and (3) radiologic imaging (ultrasonography or
computed tomography) suggesting portal hypertension
GASTROENTEROLOGY Vol. 132, No. 4
from cirrhosis. The diagnosis of cirrhosis by the ICD-9
CM code was confirmed by review of the medical records.
Patient information obtained included demographic
features; etiology of cirrhosis; laboratory test results; type
of surgical procedure; ASA class16; and dates of admission, surgery, and discharge (to determine duration of
hospitalization). Follow-up was censored at the last
known date of follow-up or date of death or liver transplantation. The study was approved by the Mayo Foundation Institutional Review Board.
Major Surgical Procedures
Patients who underwent surgery typically had
platelet counts greater than 60,000/␮L and INR of less
than 1.5 (or, before 1991, prothrombin time within
4–6 seconds of reference values) at the time of surgery.
Surgical procedures were performed with the patient
under monitored general anesthesia. Patients were divided into 3 groups by the type of surgical procedure:
digestive (n ⫽ 586), orthopedic (n ⫽ 107), and cardiovascular (n ⫽ 79). Major digestive surgery was defined as
laparotomy with operative intervention on a visceral organ. Digestive procedures were further divided arbitrarily
into 2 groups: digestive A (n ⫽ 372) and digestive B (n ⫽
214), described in Table 1.
Control Groups
Two disease control groups were included; MELD
score (median and interquartile range [IQR]) and age
(median and IQR) of each group were matched with the
major surgery group. One group, the “outpatient control
group,” included ambulatory patients with cirrhosis seen
between 2002 and 2005 without any of the above surgical
procedures (N ⫽ 562); of these, only 59 patients were
listed for liver transplantation, and 166 were not candidates for transplantation based on age or comorbid conditions. The other group (“minor surgery control group”)
included patients with cirrhosis hospitalized between
1980 and 1990 for surgical procedures under general
anesthesia, such as appendectomy, herniorrhaphy, and
other procedures considered “minor” (N ⫽ 303).4 MELD
parameters were obtained at the time of diagnosis of
cirrhosis in the outpatient group and on the day of
hospitalization in the minor surgery group. Patients
from the control groups were censored at the date of last
follow-up, death, or liver transplantation. The Charlson
comorbidity index was used to account for comorbid
conditions such as cardiopulmonary disease in all patient
and control groups.17
Calculation of the MELD Score
The MELD score was calculated from the INR,
serum total bilirubin (bili) in mg/dL, and serum creatinine (cr) in mg/dL obtained on the day of or within 24
hours before the incident operation in the surgery group
as follows6:
April 2007
CIRRHOSIS POSTOPERATIVE MORTALITY
1263
Table 1. Patient Characteristics
Groupa
Characteristic
Median age, y
Median MELD score
ASA class IV or Vf
Men
Etiology of cirrhosis
Alcoholic
Viral
Cholestatic
Other
Emergency surgery
Digestive Bc
Orthopedicd
Cardiovasculare
All
(N ⫽ 772)
Era A
(n ⫽ 233)
Era B
(n ⫽ 139)
Era A
(n ⫽ 102)
Era B
(n ⫽ 112)
Era A
(n ⫽ 50)
Era B
(n ⫽ 57)
Era A
(n ⫽ 41)
Era B
(n ⫽ 38)
63
8
142 (18)
430 (56)
61
8
55 (24)
123 (53)
64
7
6 (4)
89 (64)
62.5
8
31 (30)
46 (45)
64
7.5
14 (13)
62 (55)
63.5
7
7 (14)
28 (56)
65
6
3 (5)
24 (42)
67
10
17 (41)
31 (76)
60.5
8
9 (24)
27 (71)
182 (24)
141 (18)
142 (18)
307 (40)
79 (10)
56 (24)
28 (12)
51 (22)
98 (42)
37 (16)
26 (19)
46 (33)
17 (12)
50 (36)
2 (1)
28 (27)
16 (16)
21 (21)
37 (36)
17 (17)
14 (13)
18 (16)
22 (20)
58 (52)
13 (12)
24 (48)
7 (14)
9 (18)
10 (20)
6 (12)
14 (25)
11 (19)
13 (23)
19 (33)
0 (0)
12 (29)
8 (20)
5 (12)
16 (39)
3 (7)
8 (21)
7 (18)
4 (11)
19 (50)
1 (3)
ASA, American Society of Anesthesiologists; MELD, Model for End-stage Liver Disease.
aValues are median or number of patients (%).
bSurgery includes hepatobiliary (n ⫽ 209), portosystemic shunts (n ⫽ 86), esophagus and stomach (n ⫽ 51), and duodenum and pancreas
(n ⫽ 26). Patients in the hepatobiliary group included 82 who had hepatic resection for hepatocellular carcinoma and were reported previously.8
cSurgery includes small and large intestine (n ⫽ 170) and spleen (n ⫽ 44).
dSurgery includes knee replacement (n ⫽ 52), hip arthroplasty (n ⫽ 28), femur (n ⫽ 18), and spine (n ⫽ 9).
eSurgery includes coronary artery bypass graft (n ⫽ 27), aortic aneurysm repair (n ⫽ 18), and other/valve replacement (n ⫽ 34).
fASA class definitions: I, healthy patients undergoing surgery; II, mild systemic disease; III, severe systemic disease; IV, severe systemic disease
that is a constant threat to life; and V, moribund patients not expected to survive 24 hours with or without surgery.
MELD ⫽ 9.57 ⫻ ln(cr) ⫹ 3.78 ⫻ ln(bili) ⫹ 11.20
⫻ ln(INR) ⫹ 6.43.
Statistical Analysis
Univariate and multivariable proportional hazards models were used to determine the relationship
between preoperative variables and postoperative survival. P ⬍ .05 was considered statistically significant. The
time-dependent coefficient method of Therneau and
Grambsch was used to examine the effect of risk score
over time (eg, the effect of the MELD on 30-day, 90-day,
and 1-year mortality), and smoothing splines were used
to assess the linear relationship between continuous covariables and risk.18
Results
Demographics
Patient characteristics are shown in Table 1; follow-up was complete in all groups. The etiology of cirrhosis included alcoholic liver disease; viral hepatitis;
cholestatic liver disease (primary biliary cirrhosis and
primary sclerosing cholangitis); and other causes such as
autoimmune hepatitis, nonalcoholic steatohepatitis, and
cryptogenic. Patients in the outpatient and minor surgery
control groups did not differ from each other according
to the Charlson index (mean ⫾ standard deviation, 5.3 ⫾
3.5 and 5.6 ⫾ 3.4, respectively). The major surgery group
had a significantly higher Charlson index score (6.1 ⫾
3.6) than the outpatient (P ⬍ .001) and minor surgery
groups (P ⫽ .03). Median [IQR] age for the major surgery
group (median, 61 years [51– 69 years]) was similar to
those of the outpatient (median, 57 years [48 – 64 years])
and minor surgery (median, 56 years [47– 64 years]) control groups.
Mortality After Surgery
Patients undergoing major surgery (median [IQR]
MELD score, 8 [5–11]) had an increased risk of mortality
independent of age, MELD score, and Charlson index up
to 90 days, as compared with the outpatient (median
[IQR] MELD score, 9 [7–12]) and minor surgery (median
[IQR] MELD score, 8 [5–12]) control groups (P ⫽ .03)
(Figure 1A). Mortality at 1 year and beyond was not
different among the groups (Figure 1B and 1C).
Risk Factors for Mortality
Univariate analyses. The ASA class for patients
in the major surgery group was II in 67 patients (who
typically received a diagnosis of cirrhosis only at surgery),
III in 563 patients, IV in 132 patients, and V in 10
patients. The median survival for patients with ASA class
V was only 2 days; 9 of the 10 patients died within 2
weeks, and the other died at 85 days. These 10 patients
had a range of MELD scores (mean, 21) and ages; all
underwent emergency surgery and were excluded from
further analysis of other risk factors.
Univariate analysis of determinants for postoperative
mortality is shown in Table 2. For each variable, the effect
of time (up to 30 days, 90 days, 1 year, and after 1 year)
was examined by using time-dependent coefficient plots
(as in Figure 2). The table is thus stratified by these
intervals of follow-up. There was a significant difference
in mortality between patients with ASA classes III and IV
CLINICAL–LIVER,
PANCREAS, AND
BILIARY TRACT
Digestive Ab
1264
TEH ET AL
CLINICAL–LIVER,
PANCREAS, AND
BILIARY TRACT
Figure 1. Kaplan–Meier curves estimating mortality after surgery.
Curves show survival after surgery (major surgery patients [n ⫽ 772] and
minor surgery controls [n ⫽ 303]) or diagnosis of cirrhosis (outpatient
controls [n ⫽ 562]) through (A) 90 days, (B) 1 year, and (C) 3 years.
(hazard ratio for class IV, 2.3; (95% confidence interval,
1.9 –2.8; P ⬍ .01) but not between class III and II (hazard
ratio for class II, 0.75; 95% confidence interval, 0.5–1.1;
P ⫽ .40). The MELD score was the most important
GASTROENTEROLOGY Vol. 132, No. 4
variable in the univariate model. A single point increase
in the MELD score was associated with a 14% increase in
mortality in the first 30 and 90 postoperative days, a 15%
increase in mortality in the first postoperative year, and a
6% increase in mortality for subsequent years. The relationship between MELD score and relative risk, shown as
smoothing splines in Figure 2A–C, was almost perfectly
linear for MELD scores greater than 8; less than 8, each
point change in the MELD score had only approximately
one fourth as large an effect. From 1 to 20 years after
surgery, the excess risk was essentially constant (Figure
2D), and the relationship between MELD and increased
mortality was similar to that described in the original
model.5
Multivariable analysis. The multivariable analysis
(Table 2) was also stratified by time after surgery. Although several variables (including serum creatinine concentration, CTP score, and individual components of the
score) were significant by univariate analysis, only 3 covariables in the multivariable model—MELD score, age,
and ASA class—were significant predictors of mortality
throughout follow-up. No patients younger than 30 years
died within 90 days postoperatively; age greater than
70 years was associated with higher mortality. Male sex
added to the increased risk of mortality only during the
first year. Emergency surgery was not an independent
predictor of mortality but was the only independent
predictor of duration of hospitalization (P ⬍ .001). Median MELD score was significantly higher for those undergoing emergency surgery than for those undergoing
elective surgery (12.2 vs 7.9, respectively; P ⬍ .001).
Within the first 90 days postoperatively, the increase in
mortality related to ASA class of IV was equivalent to 5.5
MELD points; an increase in age beyond 70 years was
equivalent to 3 MELD points. The type of surgical procedure and whether it was emergency or elective were not
independently predictive of long-term mortality.
The relationship between mortality after surgery and
increasing MELD score is shown in Table 3 and Figure 3.
The median survival among all patients with digestive,
orthopedic, or cardiovascular surgery was 4.8 years for
MELD scores of 0 to 7 (n ⫽ 351), 3.4 years for scores of
8 to 11 (n ⫽ 257), 1.6 years for scores of 12 to 15 (n ⫽
106), 64 days for scores of 16 to 20 (n ⫽ 35), 23 days for
scores of 21 to 25 (n ⫽ 13), and 14 days for a MELD score
of 26 or greater (n ⫽ 10). The c-statistic for 30-day
mortality was 0.78 and for 90-day mortality was 0.82;
addition of age and ASA class improved the c-statistic
slightly to 0.80 and 0.84, respectively. These increases in
c-statistic with addition of age and ASA class were statistically significant but unlikely to be clinically important.
Discussion
In patients with cirrhosis, the risk of mortality is
related to the severity of liver disease, the comorbid
conditions, the type of surgery, and probably the skill of
April 2007
CIRRHOSIS POSTOPERATIVE MORTALITY
1265
Table 2. Univariate and Multivariable Determinants of Postoperative Mortality
Factor
Univariate analysis
MELD score ⬎8
CTP score ⬎7
Agea
ASA class IV
Surgery typeb
Cardiovascular
Digestive A
Digestive B
Orthopedic
Emergency surgery
Year of surgery
Etiology of diseasec
Alcoholic
Cholestatic
Viral/other
Male sex
Multivariable analysis
MELD score ⬎8
CTP score ⬎7
Agea
ASA class IV
Surgery typeb
Cardiovascular
Digestive A
Digestive B
Orthopedic
Emergency surgery
Year of surgery
Etiology of diseasec
Alcoholic
Cholestatic
Viral/other
Male sex
90 days
1 year
After 1 year
HR (95% CI)
P value
HR (95% CI)
P value
HR (95% CI)
P value
HR (95% CI)
P value
1.14 (1.10–1.18)
1.86 (1.03–3.37)
1.34 (1.07–1.68)
3.47 (2.05–5.86)
⬍.001
.04
.01
⬍.001
.27
.054
.16
.23
—
⬍.001
.98
.03
.30
.009
—
.04
1.14 (1.10–1.18)
2.20 (1.35–3.58)
1.38 (1.14–1.68)
4.08 (2.62–6.37)
⬍.001
.002
.001
⬍.001
.31
.06
.16
.20
—
⬍.001
.35
.01
.34
.002
—
.02
1.15 (1.12–1.18)
2.05 (1.38–3.04)
1.25 (1.12–1.40)
3.47 (2.58–4.67)
⬍.001
⬍.001
⬍.001
⬍.001
.04
.10
.005
.02
—
⬍.001
.16
⬍.001
.87
⬍.001
—
.005
1.06 (1.02–1.11)
1.21 (0.83–1.75)
1.53 (1.39–1.67)
1.84 (1.34–2.51)
.01
.32
⬍.001
⬍.001
.23
.79
.11
.33
—
.93
.02
.15
.68
.10
—
.73
⬍.001
.67
.04
.005
.60
.22
.29
.48
—
.12
.39
.27
.62
.18
—
.04
1.11 (1.06–1.16)
1.31 (0.79–2.18)
1.27 (1.05–1.54)
2.70 (1.68–4.34)
⬍.001
.30
.01
⬍.001
.79
.38
.35
.52
—
.14
.77
.16
.48
.13
—
.02
1.12 (1.09–1.16)
1.21 (0.81–1.79)
1.22 (1.08–1.37)
2.26 (1.64–3.11)
⬍.001
.35
⬍.001
⬍.001
.03
.59
.01
.08
—
.02
.009
.005
.01
.15
—
.01
1.05 (1.00–1.10)
1.08 (0.74–1.57)
1.52 (1.39–1.67)
1.47 (1.06–2.04)
3.69 (0.98–13.92)
2.35 (0.72–7.66)
2.17 (0.62–7.60)
1.00
3.20 (1.81–5.65)
1.00 (0.91–1.09)
0.73 (0.40–1.32)
0.25 (0.09–0.71)
1.00
1.79 (1.03–3.11)
1.12 (1.07–1.17)
1.14 (0.62–2.10)
1.26 (1.01–1.56)
2.21 (1.26–3.86)
2.33 (0.61–8.96)
1.90 (0.58–6.27)
1.57 (0.44–5.61)
1.00
1.72 (0.87–3.38)
0.96 (0.88–1.05)
1.19 (0.61–2.32)
0.48 (0.16–1.40)
1.00
1.79 (1.03–3.11)
2.81 (0.96–8.22)
1.95 (0.77–4.92)
1.92 (0.72–5.14)
1.00
3.09 (1.88–5.06)
1.04 (0.96–1.12)
0.78 (0.47–1.29)
0.27 (0.12–0.63)
1.00
1.76 (1.10–2.80)
1.63 (0.55–4.86)
1.56 (0.61–3.97)
1.39 (0.51–3.76)
1.00
1.54 (0.87–2.72)
0.99 (0.92–1.07)
1.23 (0.70–2.15)
0.50 (0.21–1.22)
1.00
1.74 (1.09–2.77)
1.77 (0.90–3.47)
2.14 (1.26–3.62)
1.97 (1.13–3.42)
1.00
3.07 (2.16–4.38)
0.96 (0.92–1.01)
1.03 (0.75–1.41)
0.43 (0.27–0.68)
1.00
1.51 (1.14–2.01)
1.20 (0.61–2.36)
1.95 (1.15–3.31)
1.64 (0.95–2.86)
1.00
1.55 (1.08–2.22)
0.94 (0.89–0.98)
1.49 (1.09–2.04)
0.71 (0.45–1.13)
1.00
1.45 (1.09–1.93)
1.06 (0.69–1.64)
0.77 (0.56–1.06)
0.84 (0.60–1.18)
1.00
0.98 (0.64–1.50)
1.05 (1.01–1.09)
1.06 (0.82–1.37)
0.79 (0.60–1.04)
1.00
1.04 (0.84–1.28)
1.03 (0.66–1.60)
0.92 (0.67–1.26)
0.93 (0.66–1.31)
1.00
0.71 (0.46–1.09)
1.02 (0.98–1.06)
1.12 (0.86–1.45)
1.23 (0.94–1.63)
1.00
1.05 (0.84–1.30)
.04
.68
⬍.001
.02
.91
.90
.60
.68
—
.12
.34
.31
.40
.14
—
.67
NOTE. Multivariate models include MELD score, age, ASA, and each of the other variables individually.
ASA, American Society of Anesthesiologists; CI, confidence interval; CTP, Child–Turcotte–Pugh; HR, hazard ratio; MELD, Model for End-stage Liver
Disease.
aHazard ratio for age corresponds to a 10-year increase.
bOrthopedic surgery is the reference group for procedure type.
cViral/other is the reference group for etiology.
the medical team. This study concludes that the most
important predictors of mortality are (1) severity of liver
disease as reflected by the MELD score, (2) age, and
(3) comorbid conditions as determined by the ASA physical status classification. Therefore, MELD score, ASA
class, and age may be used to characterize patients with
cirrhosis who undergo surgery according to risk of mortality, both short- and long-term. Despite advances in
anesthetic agents and surgical techniques, surgery in patients with cirrhosis today continues to be associated
with as high a risk of mortality as it was 25 years ago.
This underscores the importance of the degree of liver
dysfunction as the major determinant of postoperative
mortality in patients with cirrhosis.
The increased risk of mortality in patients undergoing
major surgery as compared with control groups may
continue for up to 90 days after surgery. The mortality
rate in the outpatient control group was higher than
previously described for patients listed for liver transplantation7 but is similar to that for cohorts of unselected patients with cirrhosis.19 The mortality rate is
lower in patients listed for transplantation because patients at highest risk for mortality undergo transplantation, thus decreasing the overall mortality rate in the
group. The stratification of risk for postoperative mortality is independent of the type of surgical procedure
performed, including whether the procedure is elective or
emergency. Of the 3 variables, ASA class V is the strongest
predictor of 7-day postoperative mortality, and MELD
score is the strongest predictor of mortality beyond 7
days and long-term. However, addition of age and ASA
class to the MELD score may further increase the predic-
CLINICAL–LIVER,
PANCREAS, AND
BILIARY TRACT
30 days
1266
TEH ET AL
GASTROENTEROLOGY Vol. 132, No. 4
CLINICAL–LIVER,
PANCREAS, AND
BILIARY TRACT
tive accuracy of determining postoperative risk of mortality. The significance of this study is that mortality risk
in the individual patient at specific time points after
surgery can now be determined (obtained by entering
the variables at: http://www.mayoclinic.org/meld/
mayomodel9.html). Thus, both the patient and physician
can make more rational decisions regarding candidacy
for surgery.
Patients with cirrhosis who undergo surgery are at risk
for mortality because of liver failure, postoperative bleeding, infection, and renal failure as a result of the circulatory dysfunction.20,21 The MELD score may be superior
to the CTP score in assessing risk of mortality because it
includes the key components of the CTP score, serum
total bilirubin (which reflects global hepatic function),
and INR (or prothrombin time), but, unlike the CTP
score, it also includes a measure of renal function, serum
creatinine. Quantitative tests of liver function such as
galactose elimination capacity and aminopyrine breath
test22 are not superior to the CTP score. The MELD score
weighs variables according to risk of mortality, has only
objective variables, and is not subject to “floor” effects or
“ceiling” effects.6 In contrast, variables in the CTP score
were selected empirically, cut-off values are arbitrary, and
similar weight is given to each of the 5 parameters in the
score. Similar to the CTP score, the MELD system has
been validated extensively in many different patient
groups with varying severity of liver disease.6,23 Whereas
the CTP score has a range only between 5 and 15 and
only 3 classes (A, B, and C), the MELD system has a much
wider range of scores, usually between 6 and 40. Therefore, a finer calibration of mortality risk is possible with
MELD. The CTP score can only help predict whether the
patient is in a low-risk, intermediate-risk, or high-risk
group, but it cannot determine the specific risk for mortality at defined time points.
Factors specific to certain surgical procedures may also
affect risk. Abdominal surgery may be associated with
decreased or impaired hepatic arterial blood flow, and
perioperative mortality rates as high as 76% have been
reported for patients with CTP class C.24 In contrast, in
patients undergoing cardiac surgery, mortality has been
related to other factors in addition to CTP class,25 such as
increased duration of cardiopulmonary bypass and need
for perioperative pressor support.26 Similar to the findings in other studies, we found that emergency surgery is
associated with a higher mortality risk, but it was not
predictive of mortality by multivariable analysis. This is
because patients undergoing emergency procedures had a
4™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™™
Figure 2. Analysis of MELD score. Smoothing spline curves illustrate
relationship between MELD score and relative risk of postoperative
mortality at (A) 30 days, (B) 90 days, and (C) 1 year. (D) Coefficient of
mortality for MELD vs postoperative time. Dotted curves represent 95%
confidence limits.
April 2007
CIRRHOSIS POSTOPERATIVE MORTALITY
1267
Table 3. Relationship Between MELD Score and Postoperative Mortality
Mortality, % (No. of patients at risk)
MELD score
7 Days
30 Days
90 Days
1 Year
5 Years
10 Years
0–7 (n ⫽ 351)
8–11 (n ⫽ 257)
12–15 (n ⫽ 106)
16–20 (n ⫽ 35)
21–25 (n ⫽ 13)
ⱖ26 (n ⫽ 10)
1.9 (314)
3.3 (236)
7.7 (94)
14.6 (29)
23.0 (7)
30.0 (6)
5.7 (301)
10.3 (219)
25.4 (78)
44.0 (19)
53.8 (4)
90.0 (1)
9.7 (287)
17.7 (200)
32.3 (69)
55.8 (15)
66.7 (3)
90.0 (1)
19.2 (253)
28.9 (170)
45.0 (56)
70.5 (10)
84.6 (2)
100 (0)
50.7 (123)
58.5 (83)
69.5 (24)
94.1 (2)
92.3 (1)
100 (0)
72.6 (57)
78.1 (35)
87.3 (10)
94.1 (2)
100 (0)
100 (0)
higher MELD score, reflecting worsening hepatic and
renal function. The advantage of the MELD score is that
it can be used to determine risk of mortality independent
of the type of surgery or intervention. Emergency surgery
does not influence long-term mortality independent of
MELD score.
Patients with ASA class V were most likely to die within
7 days after surgery. The ASA classification incorporates
measures of cardiopulmonary function that are critical to
determination of immediate survival, measures that have
been shown to be associated with a high risk of mortality
in patients with acute or chronic liver failure.27 Therefore,
postoperative mortality within 7 days may be related to
factors not reflected in the MELD score, such as cardiopulmonary comorbidity. Similarly, increasing age may be
associated with increasing cardiopulmonary comorbidity.
The series of events leading to surgery and the surgical
procedure itself may result in mortality related to liver
failure beyond 7 days, a risk best assessed by the MELD
score.
Patients with cirrhosis should have an ASA class of at
least III—that is, the presence of severe systemic illness.
Patients who were initially classified as ASA class II in this
study were those in whom cirrhosis was diagnosed only
at surgery. Therefore, it is not surprising that there was
no difference in mortality risk between ASA class II and
class III. The range of ASA classification in patients
known to have cirrhosis preoperatively (class III–V) is
small (in the present study, ASA class was III or IV in
more than 90% of the surgical patients); ASA class alone,
therefore, would allow only coarse discrimination of surgical risk over specific time periods. Moreover, there is a
subjective element in the determination of ASA class that
may limit wide applicability of the classification by less
experienced anesthesiologists. Similarly, because only extremes of age (⬍30 years or ⬎70 years) significantly
affected mortality, age alone cannot be used as a predictor of mortality in most patients. Therefore, age and ASA
class other than V must only be considered as modifiers
of the MELD score.
Because survival models typically have lower predictive
accuracy when applied to independent data sets, the
present data must be validated, especially because the
results were derived from a retrospective study. Patients
included in the study were treated at a single center by a
team of experienced surgeons, anesthesiologists, intensive care physicians, and hepatologists. Therefore, the
universal applicability of these factors in determining
postoperative risk must be determined. Moreover, there
was likely a selection bias regarding which patients underwent surgery, which depended on the perceived mor-
Figure 3. Relationship between MELD and mortality. Curves show
postoperative mortality at (A) 30 days and (B) 90 days as a function of
preoperative MELD score. The curves may be used to determine the
risk for mortality at these time points in an individual patient.
CLINICAL–LIVER,
PANCREAS, AND
BILIARY TRACT
MELD, Model for End-stage Liver Disease.
1268
TEH ET AL
CLINICAL–LIVER,
PANCREAS, AND
BILIARY TRACT
tality risk and patient preferences. The number of patients who required surgery but were not operated on
because of perceived increased risk for mortality, and the
reasons that they were denied surgery, could not be
determined in this retrospective analysis. Patients with
cirrhosis undergo surgery only if conservative measures
fail. Therefore, there is no ideal control group—that is, a
group of matched patients who require surgery but are
treated nonsurgically. Consequently, the conclusions
from the present study may not be applicable to all
patients who require surgery.
The major advantage of quantifying probability of
mortality is in making decisions regarding elective surgical procedures. Surgical procedures may be broadly classified as those that are lifesaving, such as emergency
procedures and cardiovascular and cancer surgery, and
those indicated to improve quality of life, such as orthopedic procedures. In both situations, ASA class, MELD
score, and age may be used to counsel patients and their
families regarding risk of mortality at defined times after
surgery. Mortality risk in patients with ASA class V is so
high that surgery may not be a viable option.
In patients who are candidates for liver transplantation, MELD score may be used to plan management. In
patients with a MELD score less than 11, postoperative
mortality is low enough that most patients may find the
risk of surgery acceptable (Table 3, Figure 3). Nonetheless, in our opinion, these patients should preferably have
surgery at institutions with a center for liver transplantation. In patients with a MELD score of 20 or higher, the
risk of mortality is so high that elective procedures may
be postponed until after liver transplantation. If patients
with a MELD score of 20 or higher have cardiac conditions that need correction before liver transplantation,
the cardiac surgery and liver transplantation could be a
combined procedure.26 Patients with MELD scores between 12 and 19 should have most of their evaluation for
transplantation completed before surgery so that they
can proceed to urgent liver transplantation if required.
The MELD score at which elective surgery should be
postponed until after liver transplantation may vary depending on surgical expertise and organ availability
within the center or region.
We conclude that, in patients with cirrhosis undergoing major surgical procedures, the risk of mortality
within 7 days of surgery is best assessed by ASA class,
whereas mortality after 7 days is best determined by
MELD score. Both of these variables and age may be used
to counsel patients and their families before surgery is
performed.
References
1. Rice HE, O’Keefe GE, Helton WS, Johansen K. Morbid prognostic
features in patients with chronic liver failure undergoing nonhepatic surgery. Arch Surg 1997;132:880 – 884.
2. Friedman LS. The risk of surgery in patients with liver disease.
Hepatology 1999;29:1617–1623.
GASTROENTEROLOGY Vol. 132, No. 4
3. Patel T. Surgery in the patient with liver disease. Mayo Clin Proc
1999;74:593–599.
4. Ziser A, Plevak DJ, Wiesner RH, Rakela J, Offord KP, Brown DL.
Morbidity and mortality in cirrhotic patients undergoing anesthesia and surgery. Anesthesiology 1999;90:42–53.
5. Malinchoc M, Kamath PS, Gordon FD, Peine CJ, Rank J, ter Borg
PC. A model to predict poor survival in patients undergoing transjugular intrahepatic portosystemic shunts. Hepatology 2000;31:
864 – 871.
6. Kamath PS, Wiesner RH, Malinchoc M, Kremers W, Therneau TM,
Kosberg CL, et al. A model to predict survival in patients with
end-stage liver disease. Hepatology 2001;33:464 – 470.
7. Wiesner R, Edwards E, Freeman R, Harper A, Kim R, Kamath P,
et al, United Network for Organ Sharing Liver Disease Severity
Score Committee. Model for end-stage liver disease (MELD) and
allocation of donor livers. Gastroenterology 2003;124:91–96.
8. Teh SH, Christein J, Donohue J, Que F, Kendrick M, Farnell M,
et al, Hepatic resection of hepatocellular carcinoma in patients
with cirrhosis: Model of End-Stage Liver Disease (MELD) score
predicts perioperative mortality. J Gastrointest Surg 2005;9:
1207–1215.
9. Cucchetti A, Ercolani G, Vivarelli M, Cescon M, Ravaioli M, La
Barba G, et al. Impact of model for end-stage liver disease
(MELD) score on prognosis after hepatectomy for hepatocellular
carcinoma on cirrhosis. Liver Transpl 2006;12:966 –971.
10. Suman A, Barnes DS, Zein NN, Levinthal GN, Connor JT, Carey
WD. Predicting outcome after cardiac surgery in patients with
cirrhosis: a comparison of Child-Pugh and MELD scores. Clin
Gastroenterol Hepatol 2004;2:719 –723.
11. Befeler AS, Palmer DE, Hoffman M, Longo W, Solomon H, Di
Bisceglie AM. The safety of intra-abdominal surgery in patients
with cirrhosis: model for end-stage liver disease score is superior
to Child-Turcotte-Pugh classification in predicting outcome. Arch
Surg 2005;140:650 – 654.
12. Northup PG, Wanamaker RC, Lee VD, Adams RB, Berg CL. Model
for End-Stage Liver Disease (MELD) predicts nontransplant surgical mortality in patients with cirrhosis. Ann Surg 2005;242:
244 –251.
13. Perkins L, Jeffries M, Patel T. Utility of preoperative scores for
predicting morbidity after cholecystectomy in patients with cirrhosis. Clin Gastroenterol Hepatol 2004;2:1123–1128.
14. Schroeder RA, Marroquin CE, Bute BP, Khuri S, Henderson WG,
Kuo PC. Predictive indices of morbidity and mortality after liver
resection. Ann Surg 2006;243:373–379.
15. Ji W, Li LT, Wang ZM, Quan ZF, Chen XR, Li JS. A randomized
controlled trial of laparoscopic versus open cholecystectomy in
patients with cirrhotic portal hypertension. World J Gastroenterol
2005;11:2513–2517.
16. Keats AS. The ASA classification of physical status: a recapitulation (editorial). Anesthesiology 1978;49:233–236.
17. Charlson ME, Pompei P, Ales KL, MacKenzie CR. A new method
of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis 1987;40:373–383.
18. Therneau TM, Grambsch PM. Modeling survival data: extending
the Cox model. New York: Springer, 2000.
19. D’Amico G, Garcia-Tsao G, Pagliaro L. Natural history and prognostic indicators of survival in cirrhosis: a systematic review of
118 studies. J Hepatol 2006;44:217–231. e-pub, November 9,
2005.
20. Cowan RE, Jackson BT, Grainger SL, Thompson RP. Effects of
anesthetic agents and abdominal surgery on liver blood flow.
Hepatology 1991;14:1161–1166.
21. Iwakiri Y, Groszmann RJ. The hyperdynamic circulation of chronic
liver diseases: from the patient to the molecule. Hepatology
2006;43(Suppl 1):S121–S131.
22. Gill RA, Goodman MW, Golfus GR, Onstad GR, Bubrick MP.
Aminopyrine breath test predicts surgical risk for patients with
liver disease. Ann Surg 1983;198:701–704.
23. Said A, Williams J, Holden J, Remington P, Gangnon R, Musat A,
et al. Model for end stage liver disease score predicts mortality
across a broad spectrum of liver disease. J Hepatol 2004;40:897–
903.
24. Garrison RN, Cryer HM, Howard DA, Polk HC Jr. Clarification of
risk factors for abdominal operations in patients with hepatic
cirrhosis. Ann Surg 1984;199:648 – 655.
25. Hayashida N, Shoujima T, Teshima H, Yokokura Y, Takagi K,
Tomoeda H, et al. Clinical outcome after cardiac operations in
patients with cirrhosis. Ann Thorac Surg 2004;77:500 –505.
26. Morris JJ, Hellman CL, Gawey BJ, Ramsay MA, Valek TR, Gunning
TC, et al. Case 3-1995: three patients requiring both coronary
artery bypass surgery and orthotopic liver transplantation. J Cardiothorac Vasc Anesth 1995;9:322–332.
CIRRHOSIS POSTOPERATIVE MORTALITY
1269
27. Wehler M, Kokoska J, Reulbach U, Hahn EG, Strauss R. Shortterm prognosis in critically ill patients with cirrhosis assessed by
prognostic scoring systems. Hepatology 2001;34:255–261.
Received August 25, 2006. Accepted December 14, 2006.
Address correspondence to: Patrick S. Kamath, MD, Advanced Liver
Diseases Study Group, Division of Gastroenterology and Hepatology,
Mayo Clinic, 200 First Street SW, Rochester, Minnesota 55905. e-mail:
[email protected].
Presented in abstract form at the plenary session of the American
Association for the Study of Liver Diseases (AASLD) meeting, Digestive
Diseases Week (DDW), Chicago, Illinois, May 15, 2005.
Supported in part by grant DK-34238 from the National Institute of
Diabetes, Digestive and Kidney Diseases.
The authors thank Linda Sybrant for expert secretarial assistance
and the Section of Scientific Publications, Mayo Clinic, for editing,
proofreading, and reference verification.
CLINICAL–LIVER,
PANCREAS, AND
BILIARY TRACT
April 2007