Patients with decompensated liver cirrhosis often have a limited survival, and liver transplantation is the only definite treatment modality to effectively prolong their lifespan. However, patients on the waiting list of transplantation far outnumber the potential cadaveric or living liver donors. As a result, the number of patients dying while on the waiting list is progressively increasing in recent years.1
The donor livers are allocated to the recipients according to the severity of underlying liver disease. The MELD (model for end-stage liver disease) score, which is based on the calculation from three biochemical parameters (serum bilirubin, prothrombin time, and creatinine), has been shown to more accurately predict the survival than the Child-Turcotte-Pugh (CTP) score in a recent multi-center study in the United States,2 and the liver allocation system has changed from a status-based algorithm to one using a continuous MELD severity scale to prioritize adult patients on the waiting list.3, 4 Its accuracy for outcome prediction in patients with decompensated cirrhosis has also been validated in other centers and in Europe.5–7
A major shortcoming of the currently used CTP system is that it has a “ceiling effect”, which may not be able to differentiate the disease severity among a subgroup of cirrhotic patients. For instance, patient whose serum bilirubin level is 4 mg/dL has the same CTP score as those whose bilirubin level is 20 mg/dL or higher. However, given the CTP system has this potential defect, it contains certain useful prognostic predictors such as ascites and encephalopathy. A recent study indicated that ascites is an important determinant for outcome prediction especially in those with low MELD scores.8 Other studies have suggested that inclusion of hepatic encephalopathy or hyponatremia may add additional prognostic value to the MELD score,9–11 and pointed out the potential limitation of the MELD system.12 Moreover, a concern has been raised that the CTP system was in fact not inferior to the MELD system for outcome prediction,13 and another report showed that CTP was superior to the MELD in measuring the quality of life in liver transplant candidates because ascites and encephalopathy are exclusively incorporated in the CTP system.14 To further improve, rather than to abandon, the predictive ability of the CTP system that has been used for decades, in this study we have proposed a modified CTP system with the addition of class D and investigated its performance in terms of outcome prediction.
AUC, area under the curve; CTP, Child-Turcotte-Pugh; HBV, hepatitis B virus; HCV, hepatitis C virus; INR, international normalized ratio; MELD, model for end-stage liver disease; PT, prothrombin time; ROC, receiver operating characteristic curve.
PATIENTS AND METHODS
From October 1999 to July 2004, patients who had liver cirrhosis and visited our hospital were prospectively evaluated and their medical profiles were retrospectively analyzed in this study. The following criteria were used to select eligible patients: 1) an initial CTP score of 7 or more (equivalent to class B or C) to fulfill the minimal listing criteria for liver transplantation,15 2) no coexisting hepatocellular carcinoma or human immunodeficiency virus infection at presentation, and 3) known initial MELD score and survival status at 6 months of follow-up. Four hundred and fifty patients were selected according to these criteria. Among them, 14 patients who underwent liver transplantation during the follow-up were excluded from analysis. The natural history and outcome were assessed for the remaining 436 cirrhotic patients who formed the basis of this study. None of the patients had received specific antiviral (interferon, lamivudine, or adefovir) treatment for hepatitis B or C during the study period. This study complies with the standards of Declaration of Helsinki and current ethical guidelines.
The underlying etiology of liver disease was attributed to hepatitis B virus (HBV) infection if patients were seropositive for hepatitis B surface antigen (HBsAg, RIA kits, Abbott Laboratories, North Chicago, IL) and attributed to hepatitis C virus (HCV) infection if patients were seropositive for antibody against HCV by a second-generation enzyme immunoassay (Abbott Laboratories, IL) on at least two occasions. The diagnosis of liver cirrhosis was based on the characteristic findings including physical stigmata of cirrhosis, decreased serum albumin, and increased serum globulin levels, computed tomography or ultrasonography findings of uneven liver surface, coarsened echogenicity of liver parenchyma, enlarged spleen and/or detection of ascites, and detection of esophageal varices by endoscopy. The presence and severity of ascites were detected and evaluated by computed tomography or ultrasonography. The definition of hepatic encephalopathy was according to the West Haven criteria.16
MELD and Modified CTP Score
The MELD equation used to calculate the severity score was as follows: 9.57 × loge (creatinine mg/dL) + 3.78 × loge (bilirubin mg/dL) + 11.2 × loge (INR) + 6.43.2 Minimal values are set to 1.0 for calculation purposes. The maximal serum creatinine level considered within the MELD score equation is 4.0 mg/dL.
The conventional CTP scoring system is classified from A to C and calculated on the basis of serum bilirubin and albumin levels, the prothrombin time (PT), and the presence and severity of ascites and encephalopathy. To increase the predictive ability of the CTP system, a modified scoring method was proposed and shown in Table 1 in comparison with the original CTP system. An additional 1 point was given for patient whose serum bilirubin level was > 8 mg/dL, PT prolongation > 11 sec, or albumin level < 2.3 g/dL (i.e., 5 units increase in bilirubin level, 5 units increase in PT prolongation, and 0.5 unit decrease in albumin level compared to the cutoffs used in original CTP system; referred as the “5-5-0.5 rule” for the modification of the cutoffs). These cutoffs of the laboratory parameters were so selected to fit our patient population based on the preliminary analysis in which different sets of cutoff values and possible combinations were tested to allow identifying a subgroup of patients with severe hepatic decompensation in terms of outcome evaluation. According to this score extension, a modified CTP score of 16–18 indicates severely decompensated cirrhosis and is proposed as CTP class D. All scores (MELD, original, and modified CTP) were determined at the time of referral or during the same hospitalization period.
Table 1. Proposal of a Modified Child-Turcotte-Pugh Scoring System
Original CTP class: A, 5–6 points; B. 7–9 points; C, 10–15 points.
Chi-squared test or Fisher's exact test (two-tailed) was used for categorical data. Spearman's correlation analysis was used to estimate the correlation between MELD and original CTP system, and between MELD and modified CTP system. To assess the ability of MELD and CTP score in predicting the risk of death, our analysis was performed by measuring the concordance (c-statistic) equivalent to the area under the receiver operating characteristic (ROC) curve.17 Comparison of the area under ROC curves was done using the method of Hanley and McNeil.18 The outcome was assessed as 3-month and 6-month mortality. A multivariate logistic regression analysis was used to estimate the risk ratio of mortality per 1 unit increase of the MELD, original, and modified score at 3 and 6 months. All statistical analyses were conducted using SPSS for Windows version 12 (SPSS, Inc., Chicago, IL) and MedCalc for Windows version 4.2 (MedCalc Software, Mariakerke, Belgium). For all tests, a P value less than 0.05 was considered statistically significant.
Patient Demographics and Change of CTP Scores
The baseline demographics of the study population were shown in Table 2. Patients were predominantly male and had chronic HBV infection. A substantial proportion of patients (10–28%) had a past history of esophageal varices bleeding, ascites, hepatic encephalopathy or bacterial infection at enrollment. The mean baseline MELD and original CTP score were 13.3 ± 4.2 and 8.4 ± 1.4, respectively. According to the proposed scoring method, the mean modified CTP score was 8.8 ± 1.8 (range, 7–17). The score distribution was categorized to examine the significance of the trend. There was a significant trend for the increase of the CTP score from the original to modified CTP scoring method (Fig. 1; P< 0.001). After modification of the CTP system, 100 (23%) patients had an increase of their original CTP scores for 1.5 ± 0.7 points (range, 1–3), and 94% of them had an original score of 9 or higher.
Table 2. Patient Demographics
Number of patients
66 ± 12
Etiology of cirrhosis (%)
HBsAg (+)/HBsAg (−)
Varices bleeding (%)
Bacterial infection (%)
3.0 ± 0.4
2.5 ± 1.9
INR of PT
1.2 ± 0.2
1.3 ± 0.4
Mean MELD score
13.3 ± 4.2
Mean original CTP score
8.4 ± 1.4
Mean modified CTP score
8.8 ± 1.8
Correlation of the MELD and CTP Score before and after Modification. There was a significant correlation between the MELD and original CTP score (ρ=0.509, P< 0.001; Fig. 2A). The degree of correlation appeared to be higher between the MELD and modified CTP score (ρ=0.590, P< 0.001; Fig. 2B).
Comparison of the Area Under Curve (AUC) at 3- and 6-Month Between MELD, Original, and Modified CTP Score
Of the 436 study patients, 56 (12.8%) and 93 (21.3%) patients died at 3- and 6-month respectively. Using the c-statistic and mortality taken as the end point, the AUC for modified CTP was 0.895, compared with 0.872 for the MELD (P= 0.450), and 0.809 for original CTP score (P< 0.001) at 3 months (Fig. 3A and Table 3). At 6 months, the AUC for modified CTP was 0.890, compared with 0.837 for the MELD score (P= 0.051) and 0.756 for CTP score (P< 0.001) (Fig. 3B). The predictive ability of the modified CTP was always significantly better than original CTP system and was at least equal to or better than the MELD system at different time periods by comparing the AUC. In addition, the MELD was consistently better than the original CTP system at 3- (P=0.069) and 6-month (P=0.008) for outcome prediction (Table 3).
Table 3. Comparison of the Area Under the Curve (AUC) Between MELD, Original, and Modified CTP Scoring System
P values for statistical significance: a vs. b: 0.069; a vs. c: 0.450; b vs. c: < 0.001; d vs. e: 0.008; d vs. f: 0.051; e vs. f: < 0.001.
Original CTP score
Modified CTP score
Risk of Mortality for the Increase in MELD, Original, and Modified CTP Score at 3 and 6 Months
The risk of mortality at 3- and 6-month for the MELD, original, and modified CTP system was calculated and adjusted for age, gender, and etiology of cirrhosis using multivariate logistic regression analysis (Table 4). The estimated odds ratio per unit increase of the MELD was 1.49, compared to 2.66 for original CTP and 2.70 for modified CTP at 3 months (P values all < 0.001). The corresponding odds ratio was 1.46, 2.29 and 3.08 respectively per unit increase for these three parameters at 6 months (P values all < 0.001).
Table 4. Risk of Mortality per Unit Increase in MELD, Original and Modified CTP Score at 3 and 6 Months in Multivariate Logistic Regression Model
95% Confidence interval
Modified CTP score
Original CTP score
Modified CTP score
Original CTP score
The MELD scoring system has become the prevailing criteria for organ allocation in liver transplantation.3, 19 It possesses the advantage of wide-range continuous scale to assess underlying disease severity as compared to the traditional CTP system. MELD has been shown to predict mortality across a broad spectrum of liver disease,5, 9, 20 and is useful to characterize patients with utmost need and survival benefit for a liver transplantation.21, 22 More importantly, the utilization of MELD was demonstrated to have an equal or better ability than the CTP system for short- or intermediate-term outcome prediction.2, 6, 13, 23 The application of the MELD system has also been shown a useful model to predict the outcome in cirrhotic patients undergoing major surgical procedures24 or in combination with the staging system for hepatocellular carcinoma to enhance its prognostic power.25
The CTP system and the changes in its score have been used to measure the functional dynamic liver reserve.26, 27 However, a major inherent intrinsic defect of the conventional CTP system, compared to the MELD, is that is has a ceiling effect at the highest score of 15 points.28 To overcome this flaw, we have proposed a scoring method to extend the CTP score by adding additional 1 point to patients with very high serum bilirubin level, marked PT prolongation, or with very low serum albumin level (Table 1; the “5-5-0.5 rule” for the modification of the cutoffs). Our results showed that 23% of patients would have an increase of the CTP score for 1–3 points after modification; the vast majority (94%) of them had an original score of 9 or higher. This strategy is useful to identify a subgroup of patients with severe hepatic decompensation that may have a dismal outcome, as reflected by the significant trend of the shift of the scores from the original to modified model (Fig. 1). Our results suggest that this simplified approach may greatly help refine the predictive ability of the CTP system that has been used for more than 30 years without sacrificing its basic architecture or compromising its long-standing popularity.
Most published studies have reported a MELD score's c-statistic between 0.7 to 0.9 at 3 months.12, 28 In agreement with these findings, our results showed that the AUC for MELD was 0.872 and 0.837 at 3- and 6-months, respectively. Interestingly, the AUC for the modified CTP system was even higher than the MELD at both time frames in this study (Table 3). The prognostic significance of the modified CTP system was further confirmed in the multivariate logistic regression analysis. These data support the notion that the predictive ability of the modified CTP system is not impaired by increasing its scoring categories and can be used to select patients that may have the highest mortality risk.
The modified CTP system was generated to maximize the AUC for outcome prediction in this study, and it could be argued that this approach may inflate the c-statistic and is thus unfair to compare with other models. Nevertheless, in the modified CTP system, we have only changed the cutoffs of the three laboratory-derived parameters, which are more objective and less variable. Although the modified CTP system still has a deficiency in that it does not allow for a fine resolution of patients because of fewer possible values compared to the MELD, the advantage of this strategy is that by extending the scoring categories, the variables in the CTP system can be more appropriately weighted according to their influence without the need for logarithmic transformation, and the modified system is able to differentiate disease severity and improve its performance by partially offsetting the ceiling effect.
A distinct feature of this study, however, is that the majority of patients had chronic HBV infection which is prevalent in many Asian countries. Therefore, a major limitation of this study is that our findings may not be readily applicable to the US or Europe where hepatitis C and alcoholism are common etiologies of end-stage liver disease. Disease progression could be highly variable depending on the status of HBV replication, and more studies are needed to clarify the potential impact of viral factor on the outcome.
The application of the MELD has raised some controversial issues. Recent studies suggested that certain variables, such as serum sodium level,10, 11 ascites,8 encephalopathy,9 cirrhosis-related complications,29 and development of hepatocellular carcinoma30, 31 may also be important prognostic predictors that are not included in the MELD. In addition, the change of MELD score over time has been suggested as a marker to predict the survival in cirrhotic patients awaiting liver transplantation.32–34 While MELD is more objective, repeatable, and verifiable and is superior to the conventional CTP for comparison at a population level, it can be anticipated that the CTP system will resist as an intuitive score for bedside assessment. Alternatively, since not all scoring systems designed for patient care need to be applicable in an allocation system, the modified CTP system could be more suitable for local health care system because it is still partly based on subjective measures.
In conclusion, our results indicate that modified CTP system is a valuable model in predicting the short and intermediate term outcome for cirrhotic patients. By overcoming the ceiling effect of the original CTP system, the modified CTP system greatly improves its predictive ability and is as efficient as the MELD system in terms of outcome prediction. We propose the modified CTP system can be used as an alternative prognostic model for cirrhotic patients awaiting liver transplantation. Further studies based on actual outcome data are needed to confirm our findings.