Laboratory test variability and model for end-stage liver disease score calculation: effect on liver allocation and proposal for adjustement.

BACKGROUND: The use of the Model for End-Stage Liver Disease (MELD) score to prioritize patients on liver waiting lists must take the bias of different laboratories into account. METHODS: We evaluated the outcome of 418 patients listed during 1 year whose MELD score was computed by two laboratories (lab 1 and lab 2). The two labs had different normality ranges for bilirubin (maximal normal value [Vmax]: 1.1 for lab 1 and 1.2 for lab 2) and creatinine (Vmax: 1.2 for lab 1 and 1.4 for lab 2). The outcome during the waiting time was evaluated by considering the liver transplantations and the dropouts, which included deaths on the list, tumor progression, and patients who were too sick. RESULTS: Although the clinical features of patients were similar between the two laboratories, 36 (13.1%) out of 275 were dropped from the list in lab 1, compared to 5 (3.5%) out of 143 in lab 2 (P<0.01). The differences were mainly due to the deaths on the list (8% lab 1 vs. 2.1% lab 2, P<0.05). The competing risk analysis confirmed the different risk of dropout between the two labs independently of the MELD score, blood group, and preoperative diagnosis. The bias on MELD calculation was considered and bilirubin and creatinine values were "normalized" to Vmax of lab 1 (corrected value=measured value x Vmax lab 1/Vmax lab 2). By comparing receiver operating characteristic curves, the ability of MELD to predict the 6-month dropouts significantly increased from an area under the curve of 0.703 to 0.716 after "normalization" (P<0.05). CONCLUSIONS: Normalization of MELD is a correct and good compromise to avoid systematic bias due to different laboratory methods.