Child-Pugh vs MELD: which liver score to use

How Child-Pugh and MELD are built

The Child-Pugh score and the MELD score measure liver disease severity in different ways. Child-Pugh adds 1 to 3 points for each of five items: hepatic encephalopathy, ascites, bilirubin, albumin and INR (prothrombin time in Pugh's 1973 table), for a total of 5 to 15 and a class of A (5 to 6), B (7 to 9) or C (10 to 15). Two of the five items, ascites and encephalopathy, are graded by the clinician. The Model for End-Stage Liver Disease (MELD) is a formula built from three lab values only: bilirubin, creatinine and INR, entered as natural logarithms, which gives a continuous score from 6 to 40. MELD-Na adds serum sodium, and MELD 3.0 adds sodium, albumin and female sex. So Child-Pugh includes clinical signs and albumin but no kidney function, while MELD includes kidney function but, until MELD 3.0, neither albumin nor any clinical grade.

Where each score came from

Child-Pugh and MELD were both built to predict risk around a procedure, decades apart. Child and Turcotte graded patients before shunt surgery, and in 1973 Pugh and colleagues modified that grading for patients having oesophageal transection for bleeding varices, dropping nutritional status and adding prothrombin time. MELD began as the Mayo TIPS model: Malinchoc and colleagues studied 231 patients having elective transjugular intrahepatic portosystemic shunts (TIPS) and found that bilirubin, creatinine, INR and the cause of liver disease predicted survival better than the Child-Pugh class or score. In 2001 Kamath and colleagues tested the model in four separate groups and found it predicted death within 3 months with c-statistics of 0.78 to 0.87, while complications of portal hypertension such as ascites added at most 0.03. That objective, lab-only design is why the US Organ Procurement and Transplantation Network (OPTN) chose MELD to rank liver transplant candidates.

Which predicts mortality better?

MELD predicts short-term death slightly better in transplant candidates, but across all settings the two scores perform similarly. Wiesner and colleagues applied MELD to 3,437 adults added to the US liver transplant waiting list from November 1999 to December 2001, of whom 412 (12%) died within 3 months. The c-statistic for 3-month mortality was 0.83 for MELD against 0.76 for the Child-Turcotte-Pugh score. A broader answer comes from the 2016 meta-analysis by Peng, Qi and Guo, which found 119 studies with 269 comparisons of the two scores: 44 favoured MELD, 16 favoured Child-Pugh, 99 found no significant difference and 110 did not report significance. In subgroups, Child-Pugh was more sensitive but less specific in acute-on-chronic liver failure, MELD was more sensitive in patients admitted to intensive care, and Child-Pugh was more specific in patients having surgery. The authors concluded that the scores have similar prognostic value in most cases.

Why the two scores can disagree

Child-Pugh and MELD disagree most when a patient's problem falls in the part of the liver picture that only one score sees. Child-Pugh caps every item at 3 points, so a bilirubin of 4 mg/dL and one of 30 mg/dL score the same, while MELD keeps rising with each value up to its maximum of 40. Kidney injury raises MELD sharply through creatinine but leaves Child-Pugh unchanged. Ascites and encephalopathy drive Child-Pugh up but are absent from MELD, and Kamath's data showed that ascites added little to MELD's 3-month prediction. Low albumin scores up to 3 Child-Pugh points but counts in MELD only from version 3.0. Treatment can move Child-Pugh too: diuretics and paracentesis lower the ascites grade and infused albumin raises the albumin value. The two examples below show how far apart the scores can sit for the same day's results.

Patient A has severe decompensation by Child-Pugh but a modest MELD, because her labs, apart from albumin, are near normal. Patient B is class A, yet his kidney function puts his MELD-Na in the 20 to 29 tier, where Wiesner's cohort had a 3-month mortality of 19.6%. You can check both with the Child-Pugh calculator and the MELD calculator.

Which score to use, and when

Use MELD when the question is short-term mortality or transplant priority, and Child-Pugh when you need a widely understood description of how well the liver is working. MELD is the score transplant systems use: OPTN ranked candidates aged 12 and over by MELD-Na from January 2016 and switched to MELD 3.0 in July 2023, because a calculated score avoids the observer variation of graded ascites and encephalopathy. Child-Pugh remains the common language of cirrhosis in clinic notes, surgical risk discussions, hepatocellular carcinoma care and trial eligibility, where class A, B or C is quicker to communicate than a number. Many clinicians record both, because each catches what the other misses: Child-Pugh shows the decompensation that MELD ignores, and MELD shows the kidney injury and extreme lab values that Child-Pugh caps or leaves out. Neither replaces clinical assessment, and both should be repeated as the patient's condition changes.