Mortality risk score in liver transplantation: changes over time in its predicting power

Since the onset of our liver transplantation program in 1992, 362 transplants were performed in 338 children. A risk score for predicting mortality was designed and implemented over time. The description of a method utilized to design the risk score, changes in mortality rate over 12 yr and the analysis of factors that might have influenced these changes are presented and discussed in this paper.

Cox regression analysis was applied to a retrospective sample of 110 patients with liver cirrhosis, transplanted between 1992 and 2000. A risk score was prepared using beta coefficients of the two significant variables related to survival time: age (1.08, p=0.02) and bilirubin levels (0.93, p=0.03), and two groups were identified: low- and high-risk score. The score was applied in two consecutive samples: 2000-2002 and 2002-2004.

In the first sample (1992-2000), we found 69 and 41 as low- and high-risk patients, with a median survival time of 93.13 and 2.93 months (p=0.0001). In the 2000-2002 sample, a median survival time of 41.7 and 2.33 months (p=0.03) was found for low- and high-risk groups, respectively. In the third sample (2002-2004), there was a remarkable decrease in mortality in the high-risk group (n=29) and the score did not discriminate between high- and low-risk groups (p=0.35).

A scoring system to identify risk levels in liver transplantation patients is an operative and powerful tool during a given period of time but it has to be updated as risk factors will vary following the team's learning curve.