Assessment of Prognosis of Cirrhosis

François Durand, M.D.; Dominique Valla, M.D.


Semin Liver Dis. 2008;28(1):110-122. 

In This Article

Meld Score

While Child score was originally designed for assessing the prognosis of cirrhotic patients undergoing surgical treatment of portal hypertension, MELD score was designed for assessing the prognosis of cirrhotic patients undergoing transjugular portosystemic intrahepatic shunt (TIPS).[24] Multivariate analysis using Cox regression analysis showed that among a list of pre-determined variables, four variables had an independent impact on survival, namely bilirubin, creatinine, INR, and the cause of cirrhosis (alcoholic and cholestatic versus other causes). To lessen the influence of extreme values, the natural logarithm of bilirubin INR and creatinine were entered into the model. Based on regression analysis, a coefficient was attached to each variable, according to the weight of each variable on mortality risk. In the original series, the resulting score was slightly more accurate than Child-Pugh score for predicting survival after TIPS.

With the expansion of liver transplantation, attention moved from the management of portal hypertension to waiting list mortality and organ allocation policy in patients listed for transplantation. Indeed, waiting time alone proved to be an inaccurate marker of waiting list mortality[27] and more relevant markers were needed. A slightly modified risk score, termed the MELD score, was tested in populations of cirrhotic patients for assessing early (3 months) mortality risk after placement on the waiting list.[28] This modified score proved to be a robust marker of early mortality which could be generalized to patients with various causes of cirrhosis and various degrees of severity. Excluding the variable “cause of cirrhosis” had a minimal impact on the model accuracy. As a result, a simplified version of MELD score including only three objective variables (bilirubin, creatinine, and INR) was eventually proposed for easier use ( Table 3 ).[29] According to this modified score, patients with bilirubin and creatinine values below 1 mg/dL (17 and 90 μmol/L, respectively) are rounded off to 1 mg/dL to avoid negative logarithmic values. Similarly, patients with INR below 1 are rounded off to 1. Whatever the individual values, the score is empirically capped at 40. Consequently, MELD score represents a continuous variable ranging from 6 to 40.

MELD score has been adopted since 2002 for organ allocation to patients listed for liver transplantation in the United States.[30] According to the MELD-based policy, patients with the highest score have a priority for organ allocation. More recently, MELD score has also been adopted in several European countries as well as in South America. The impact of MELD score on liver transplantation is discussed below.

In addition to organ allocation, several studies have confirmed that MELD score is a reliable tool for predicting outcome after TIPS.[31,32] The “c” statistic represents a global estimate of the ability of a score to predict an event. This statistic, which is derived from the area under the receiver operating characteristic curve, ranges from 0 to 1. A “c” statistic of 0.5 means that the score is of no value for predicting a given event. A “c” statistic of 1 means that the score is perfect (a goal which is never achieved in clinical practice). The “c” statistic of the MELD score for predicting 1-year survival after TIPS is ∼0.7, which means that it is clinically useful. MELD score also proved to be a reliable marker of 1-year and 5-year survival across a broad spectrum of liver diseases including alcoholic cirrhosis and alcoholic hepatitis.[33] In addition, MELD score has been shown to be a good prognostic marker in cases of variceal bleeding,[34] spontaneous bacterial peritonitis,[35] and hepatorenal syndrome.[36] Independent of the cause of cirrhosis, high MELD score was shown to be associated with a decrease in residual liver function as measured by monoethylglycinexylidine test.[37] The particular issues of alcoholic cirrhosis, HCC, nontransplant surgery, and transplantation are detailed below.

Using statistical analysis rather than empirical selection of variables for establishing a prognostic score can be anticipated as more rigorous, on a methodological basis, and eventually, more efficacious. However, it must be kept in mind that even if the three components of the MELD score have been selected on the basis of a statistical analysis, all the variables entered in the model have been empirically selected, because they were felt to have an influence on the outcome or, more simply, because they were available. It cannot be excluded that other important variables may not have been taken into account.

The three variables entered in the MELD score (bilirubin, creatinine, and INR) are supposed to be objective (in contrast to ascites and encephalopathy, the grading of which is subjective). However, even these “objective” markers may be subjected to significant variations resulting from either changes over time or from different laboratory methodologies. There are substantial interlaboratory variations in INR depending on the methods used for determination. It has been shown that interlaboratory variation in INR is ∼25%. Among the three variables of MELD score, INR has the highest multiplicative value. Therefore variations in INR may translate to up to 20% differences in MELD score.[38] In single individuals with cirrhosis, substantial changes in serum creatinine may occur, especially in those undergoing large-volume paracentesis and/or receiving diuretics. Laboratory methods may also interfere with the value of serum creatinine. There is a poor agreement among different creatinine assays, especially as serum bilirubin rises.[39] Overall, MELD score is not as objective as it was expected to be.

A major limitation of MELD score is the need for computation, which makes it less friendly to use than Child-Pugh score at the bedside. Logarithmic transformation has been chosen to optimize the statistical model. Originally, this statistical model and the derived score were not designed to be routinely used in clinical practice. With the expansion of MELD score in many fields of hepatology, the need for computation represents a source of difficulties. There is no evidence that a simplified score based on the original values of bilirubin, INR, and creatinine (without logarithmic transformation) would be less accurate. In addition, there are no clear-cut values of MELD score for easily categorizing individual patients according to their own mortality risk.


Comments on Medscape are moderated and should be professional in tone and on topic. You must declare any conflicts of interest related to your comments and responses. Please see our Commenting Guide for further information. We reserve the right to remove posts at our sole discretion.