Model for end-stage liver disease (MELD) score is a good parameter to establish the patient survival before liver transplantation and give priority to the sickest patients. The aim of this study was to evaluate the variability and potential regression of MELD score during the months before liver transplant. From the 350 patients waitlisted for transplantation, we evaluated the 124 patients who had enough blood tests during 12 months before the final event (transplantation, death, removal from list due to improvement or worsening). We considered month 12 as the final event and blood tests from 0, 3, 6, and 12 months were analyzed. MELD score was calculated and compared using ANOVA for repeated measures test. To determine variability of MELD and its components, intraclass correlation coefficient (ICC) was calculated for 0, 3, and 6 months. The degree of constancy was defined by proximity of ICC to 1. Two groups by initial MELD (< or =17 or >17) were considered. Patient data are: mean age, 53 +/- 9 years; sex: 70% men, etiology, 28% hepatitis C, 11% alcohol and hepatitis C, 16% alcohol, 28% hepatocellular carcinoma, 6% hepatitis B, 11% others; Initial Child-score, 8.5 +/- 2.0; Initial MELD score, 15.2 +/- 4.9; mean time on waiting list, 8.1 +/- 5.7 months. MELD score from 6 and 12 months was significantly higher than the initial one. The most constant parameter was creatinine (ICC:0.89); bilirubin (ICC:0.58) and INR (ICC:0.59) were the most variable ones. MELD score ICC was 0.79. In only one patient did MELD score decrease 5 points below the initial one. For initial MELD < or = 17 and >17, variability was lower in the former. In conclusion, MELD became significantly higher 6 months after the basal determination. This score is reliable as it does not tend to decrease in time. In high MELD scores (>17), 3-month survival was lower and variability greater so that more careful follow-up and prioritizing are needed.