Normalised intrinsic mortality risk in liver transplantation: European Liver Transplant Registry study

Background No model exists for liver transplantation to estimate the mortal ity risk in a given patient, and no standard by which to assess performance in different centres. We investigated the intrinsic mortality risk in the absence of known mortality risk factors, Methods We identified mortality risk factors and risk ratios quantified in data from the European Liver Transplant Registry (22 089 patients at 102 ce ntres in 18 countries) registered from 1988 to 1997. To develop a model of the intrinsic risk and the risk ratios for specific factors, univariate and multivariate analyses were done separately for the overall population, for adults, and for children younger than 15 years, and the number of deaths w ere estimated, We validated the model by comparing mortality in patients wi thout risk factors with the model-adjusted mortality in patients with risk factors. Findings Overall 5-year and 8-year actuarial survival was 66% (95% CI 65-66 ) and 61% (60-62). 65% of deaths occurred within 6 months, Retransplantatio n, transplantation for cancer, acute liver failure, fewer than 20 split-liv er grafts per year, and a centre workload of fewer than 25 transplants per year were the main risk factors of 12 identified factors, 1-year and 5-year death rates among adults with no risk factors were similar to model estima tes (15 [13-16] vs 14% [13-15], and 22 (20-24) vs 23% [21-24]). Correspondi ng data for paediatric transplants were 9% (7-12) compared with 11% (9-12) and 13% (10-17) compared with 14% (11-16). The reduction of mortality risk in high-volume centres was even greater in patients without risk factors (4 8 vs 23%, p<0.001). Interpretation The normalised intrinsic mortality risk can be combined with the relative risk ratios of known risk factors to better estimate the mort ality risk of a given procedure in a given patient. Centres can assess perf ormance by removing potential bias of donor and recipient selection.