The perioperative outcome of hepatic resection on cirrhosis has improved remarkably in recent years with improved surgical techniques and perioperative care.1–4 Even so, proper preoperative assessment of liver function reserve remains essential to ensure a low incidence of postoperative liver failure and mortality.
Several approaches have been proposed to assess hepatic function reserve, but of all the tools, Child-Turcotte-Pugh (CTP) classification5, 6 remains the most widely accepted system that provide an initial clue to the extent of resection that a cirrhotic patient can tolerate.7–9 CTP Class C cirrhosis is considered an absolute contraindication for hepatic resection in most of the major hepatobiliary centers and only minor resection would be considered for CTP Class B cirrhotic patients. For patients with CTP Class A cirrhosis, the decision for hepatectomy often requires additional liver function tests that provide more refined evaluation of liver function reserve.10–13
Recently, the model for end-stage liver disease (MELD) was developed to predict mortality of cirrhotic patients receiving transjugular intrahepatic portosystemic shunts14 and subsequently applied as a disease severity index for priority on the waiting list for liver transplantation.15, 16 Thus, the MELD score can be considered as an index of hepatic functional reserve, but its ability to predict liver failure after resection has been never assessed.
The aim of this study was to retrospectively establish whether the MELD score can predict postoperative liver failure after hepatectomy for hepatocellular carcinoma (HCC) in cirrhotic patients. Its relationship with postoperative complications, hospital stay, 1-year survival, were also investigated.
CTP, Child-Turcotte-Pugh; HCC, hepatocellular carcinoma; MELD, model for end-stage liver disease.
PATIENTS AND METHODS
Between January 1997 and December 2004, 229 patients underwent curative hepatic resection for HCC on chronic liver disease at the Department of Surgery and Transplantation of the University of Bologna. The policy of our center regarding indications for hepatectomy has been published elsewhere.1 Of these 229 patients, 32 were not included in the analysis for the following reasons: incomplete clinical data (27 patients), presence of chronic renal insufficiency that did not allow a reliable measure of the MELD score (3 patients) and early postoperative death due to acute myocardial infarction (2 patients).
The initial study group had 197 patients. Hepatitis activity and fibrosis stage were scored according to the classification of Ishak et al.17: Necrosis and inflammatory changes characteristic of hepatitis were scored as mild (0-5), moderate (6-12), or severe (13-18), and fibrosis was scored as cirrhosis vs. noncirrhosis. Forty-three patients (21.8%) had underlying, histologically proven, mild to severe fibrosis without cirrhosis leading to the exclusion of this group from the analysis. The final study group had 154 patients.
Table 1 reports the clinical and pathological features collected for each patient. CTP score was obtained according to classification proposed by Pugh.6 The MELD score was calculated according to the following formula: MELD = (0.957 × Loge [creatinine in mg/dL] + 0.378 × Loge [bilirubin in mg/dL] + 1.12 × Loge [INR] + 0.643) × 10.16
Table 1. Baseline Characteristics of the Study Population
Patients in Study (n = 154)
NOTE: Continuous variables are reported as medians and ranges.
Underlying liver disease
Hepatitis B + C
Hepatitis non-B and non-C
Grade of hepatitis activity
Serum creatinine (mg/dL)
Serum bilirubin (mg/dL)
International normalized ratio
Extension of hepatectomy
Size of the largest tumor (cm)
Major hepatic resection was defined as the removal of more than 2 segments. Extension of hepatectomy was referred to International Hepato-Pancreato-Biliary Association (IHPBA) classification.18 All resections were performed with a tumor-free margin of at least 1 cm confirmed by the pathologist.
Tumor histological grading was assessed according to the criteria of Edmondson and Steiner based on the areas of the tumor having the highest grade.19 Tumors were classified using the pathologic tumor-node metastases previously proposed.20
The primary end point of the study was the investigation of the relationship existing between the preoperative MELD score and the development of irreversible liver failure after hepatectomy of cirrhotic patients. It was defined as a growing impairment of liver function after resection that led to patient death or required transplantation.
The relationship between the MELD score and postoperative complications (morbidity), length of hospital stay, and 1-year patient survival represented secondary end points. Postoperative morbidity included occurrence of refractory ascites (requiring drainage), jaundice (defined as serum bilirubin level above 5 mg/dL, corresponding to 294 μmol/L), severe alteration of coagulation factors requiring fresh frozen plasma infusion, and renal impairment (defined as an increase of blood urea nitrogen above 2 g/L and/or an increase of serum creatinine above 2 mg/dL, corresponding to 182 μmol/L).
Hospital stay was computed from the day of surgery until discharge at home. Patient survival was computed from the day of surgery until the most recent follow-up. Controls and patients still alive after the first year after surgery were censored at this time point; patients transplanted for postoperative liver failure were censored the day prior to transplantation, and patients dead for causes not related to liver failure were censored the day prior to the event.
Normal distribution could not be proved for the clinical variables available (Kolmogorov-Smirnov test, P < 0.05) and nonparametric tests were applied to all the data analysis. Continuous variables were expressed as median and range, and the values in the different subgroups were compared using the Kruskal-Wallis test. Categorical variables were expressed as prevalence and subgroups were compared using the chi-square test with Yates's correction. Survival probabilities were constructed using Kaplan-Meier survival estimates and compared using the log-rank test.
The prognostic value of MELD in predicting postoperative liver failure and complications was assessed using receiver operating characteristic (ROC) curve analysis.21 The area under the curve (AUC) and the sensitivity and the specificity for cutoff points obtained were reported.
A significance level of 0.05 was used in all analyses. The statistical analysis was done using SPSS Version 10.0 software (SPSS, Chicago, IL,), and ROC analysis was performed using MedCalc Version 18.104.22.168 (MedCalc Software, Mariakerke, Belgium).
The study cohort included 112 men (72.7%) and 42 women (27.3%). The median age was 64 years (range, 41-81) (Table 1). One hundred forty-three patients were classified as CTP Class A (92.9%) and 11 patients as CTP Class B, Score 7 (7.1%); median MELD score prior to surgery was 9 (range, 6-15). Patients with CTP Class A cirrhosis showed a median MELD score of 8, ranging from 6 to 14, significantly lower than patients with CTP Class B cirrhosis who had a median MELD score of 11 (range, 9-15, P = 0.001).
Surgery consisted of minor hepatic resections in 146 cases (94.8%) and of major hepatic resection in the remaining 8 (5.2%). In particular, 74 wedge resections (48.1%), 49 segmentectomies (31.8%), 11 left lateral sectionectomies (7.1%), 5 right posterior sectionectomies (3.2%), 4 right anterior sectionectomies (2.6%), 3 left medial sectionectomies plus segmentectomy 8 (1.9%), 6 left hepatectomies (3.9%), and 2 right hepatectomies (1.3%) were performed.
Eleven patients (7.1%) developed irreversible postoperative liver failure. Of these, 7 patients classified as CTP Class A underwent a wedge resection in 3 cases, a right posterior sectionectomy in 2 cases and a left hepatectomy in 2 cases; conversely, 4 patients classified as CTP Class B underwent a wedge resection in 3 cases and a segmentectomy in the remaining 1 case. Patients who experienced postoperative liver failure had a median MELD score of 11 (range, 9-15), significantly higher in comparison to patients in which this event did not occur (median, 8; range, 6-13; P = 0.001). ROC analysis identified a MELD score equal to or above 11 as the best cut-off value for predicting postoperative liver failure (AUC = 0.92, 95% CI = 0.87-0.96; sensitivity = 82%; specificity = 89%) (Fig. 1A).
Forty-six patients (29.9%) experienced at least 1 postoperative complication. Refractory ascites developed in 40 cases (26%), jaundice in 31 cases (20.1%), alteration of coagulation factors in 34 cases (22.1%), and renal impairment in 10 cases (6.5%). Patients who experienced postoperative complications had higher MELD score (median, 10; range, 7-15) in comparison to patients who did not experience any complication (median, 8; range, 6-11; P = 0.001). ROC analysis identify MELD score equal to or above 9 as the best cutoff value for predicting occurrence of postoperative complications (AUC 0.85, 95% CI = 0.78-0.89; sensitivity = 87%; specificity = 63%) (Fig. 1B). Median hospital stay of cirrhotic patients was 9 days (range, 5-166) and 1-year survival was 93%.
Patients were divided according to the cut-offs of the MELD scores obtained by ROC analysis in 3 groups: MELD below 9, MELD between 9 and 10, and MELD equal to or above 11. Seventy-four patients (48%) had a MELD score below 9, 56 patients (36.4%) had a score between 9 and 10, and 24 patients (15.6%) had a score equal to or above 11. Table 2 reports the prevalence of postoperative liver failure and the morbidity in relationship with MELD score prior to surgery; median hospital stay and 1-year survival rates were also reported. Patients with a MELD score below 9 did not experience postoperative liver failure and showed the lowest morbidity (8.1%); patients with a score between 9 and 10 showed an increase of postoperative liver failure (3.6%) and morbidity (35.7%); and patients with a score equal to or above 11 had the highest prevalence of postoperative liver failure (37.5%) and the highest morbidity (83.3%) (P = 0.001 in both cases).
Table 2. Relationship of MELD Score Prior to Surgery and Postoperative Irreversible Liver Failure, Morbidity, Hospital Stay and 1-Year Patient Survival
Comparison among patients with low, intermediate, and high MELD scores.
Postoperative liver failure
Alteration of coagulation factors
Hospital stay (days)
Patients with a MELD score below 9 were discharged from the hospital after a median of 8 days (range, 5-38) and were all alive within 1 year of surgery; patients with a score between 9 and 10 showed a significant increase of the hospital stay (median, 9 days; range, 6-33) and a decrease of 1-year survival (94%), whereas the longest hospital stay (median, 25 days; range, 6-166) and the worst 1-year survival (74%) were observed in patients with MELD score equal to or above 11 (P = 0.001 in both cases) (Table 2).
Postoperative liver failure is one of the most dreadful complications of liver resection, and it is even more dreaded in patients with liver cirrhosis, because resection removes functional liver tissue from an organ that is already functioning marginally. Preoperative assessment of liver function and prediction of postoperative functional reserve are of paramount importance to minimize surgical risk.
The present study demonstrated that MELD score can accurately predict both postoperative liver failure (AUC = 0.92) and postoperative morbidity (AUC = 0.85) of cirrhotic patients undergoing resection for HCC. A MELD score equal to or above 11 prior to surgery lead to a very high incidence of liver failure (37.5%) and postoperative complications (83.3%), resulting in a significantly longer hospital stay and a lower 1-year survival in comparison to patients with lower MELD scores.
Providing a more accurate partition of CTP Class A patients, MELD assessment may identify those patients who are at high risk of postoperative liver failure and those who may be safely treated with hepatectomy: a patient eligible for resection on the basis of CTP score may be not resectable on the basis of MELD. These patients should be referred to nonsurgical approaches such as radiofrequency ablation or transarterial chemoembolization. Conversely, patients with MELD scores below 9, who represent the 48% of cirrhotic patients of the present study, showed no postoperative liver failure, as well as low morbidity, shorter hospital stays and a 100% survival rate, proving that a good outcome can be achieved in these patients through hepatic resection.
The possibility of having a reliable index of resectability remains one of the main challenges for many researchers. Several quantitative tests have been proposed in the evaluation of liver function of patients with HCC referred for resection; however, CTP score remains the main indicator for determining eligibility for surgical treatment.6 Many studies have subsequently shown that CTP index can be useful in the assessment of hepatic functional reserve in cirrhotic patients referred for resection.7–9 However, the decision for hepatectomy often requires additional liver function tests that provide more refined evaluation of liver function reserve namely indocyanine green clearance test,10, 11 lidocaine test,12 or galactose elimination capacity13 as a guidance for selection of optimal patients.
Similar to the CTP score, the MELD score was originally developed to predict survival of cirrhotic patients undergoing transjugular intrahepatic portosystemic shunt and then used as a disease severity index determining priority on the waiting list for liver transplantation.14–16 Thus, the MELD score is considered an accurate index of hepatic functional reserve, and it has been reported to accurately predict postoperative morbidity and mortality of cirrhotic patients undergoing abdominal surgery.22, 23 However, at present and to the best of our knowledge, this is the first report that evaluate its impact on prognosis of hepatectomy for HCC on cirrhosis.
Several reports from different groups, including ours, consider partial hepatectomy a good option in well-compensated CTP Class A patients.1, 24–26 Otherwise, liver resection has been considered as a bridge to liver transplantation in order to reduce the dropout rate in patients with HCC on cirrhosis on the waiting list.27 Patients with a MELD score equal to or above 11 probably represent the ideal patients to refer for nonsurgical approaches such as thermal ablation or chemoembolization and, whenever possible, for liver transplantation. In particular, in the setting of liver transplantation, Merion et al.28 showed that the threshold score of 11 patients in whom, independent of HCC, transplantation is justified (above 11) or futile (below 11); in particular, cirrhotic patients with a MELD score between 6 and 11 were shown to have a posttransplant mortality significantly higher than waiting list mortality. Therefore, the results from the present study combined with those of Merion et al. may give support to a systematic transplantation policy in small HCC patients with a MELD score exceeding 11 as well as partial hepatectomy in patients with lower MELD scores. Even in patients with MELD score between 9 and 10, strict care must be taken in intraoperative and postoperative management and major hepatectomies should be avoided.
In conclusion, the MELD score can accurately predict liver failure and morbidity in patients with HCC on cirrhosis undergoing hepatic resection. Cirrhotic patients with a high MELD score are at high risk of postoperative liver failure and complications and should be referred for other treatments. Cirrhotic patients with low MELD scores treated with minor hepatic resections achieve no mortality and low morbidity rates. Application of the MELD score in the preoperative assessment of liver function prior to hepatic resection is recommended.