ABSTRACT
Current diagnostic criteria for primary nonfunction (PNF) of liver grafts are based on clinical experience rather than statistical methods. A retrospective, single-center study was conducted of all adults (n = 1286) who underwent primary liver transplant (LT) 2000-2008 in our center. Laboratory variables during the first post LT week were analyzed. Forty-two patients (3.7%) had 2-week graft failure. Transplant albumin, day-1 aspartate aminotransferase (AST), day-1 lactate, day-3 bilirubin, day-3 international normalized ratio (INR), and day-7 AST were independently associated with PNF on multivariate logistic regression. PNF score =(0.000280*D1AST)+ (0.361*D1 Lactate)+(0.00884*D3 Bilirubin)+(0.940*D3 INR)+(0.00153*D7 AST)-(0.0972*TxAlbumin)-4.5503. Receiver operating curve analysis showed the model area under receiver operating curve (AUROC) of 0.912 (0.889-0.932) was superior to the current United Kingdom (UK) PNF criteria of 0.669 (0.634-0.704, p < 0.0001). When applied to a validation cohort (n = 386, 34.4% patients), the model had AUROC of 0.831 (0.789-0.867) compared to the UK early graft dysfunction criteria of 0.674 (0.624-0.721). The new model performed well after exclusion of patients with marginal grafts and when modified to include variables from the first three post-LT days only (AUROC of 0.818, 0.776-0.856, p = 0.001). This model is superior to the current UK PNF criteria and is based on statistical methods. The model is also applicable to recipients of all types of grafts (marginal and nonmarginal).