Glasgow coma score and tumor necrosis factor α as predictive criteria for initial poor graft function

Initial poor graft function (IPGF) is a major factor influencing the clinical outcome after liver transplantation (LT), but there is no reliable method to assess and predict graft dysfunction. To help clinicians determine prognosis in the early postoperative period, individual parameters and complex scoring systems have been suggested, but most of them are inaccurate because of the multifactorial nature of transplantation courses. Therefore, the aim of our study was to retrospectively evaluate predictive criteria for retransplantation. Forty-two patients were enrolled in this study: 18 who experienced primary non-function (PNF) and 24 with delayed graft function (DGF). All of the patients were treated with the Molecular Adsorbent Recirculating System (MARS). They were into 3 subgroups: patients who survived without LT (n = 20; 47.7%); patients who underwent LT (n = 16; 37%), and patients who died before transplantation (n = 6; 14%). Stepwise multivariable logistic regression analysis was performed with the intent to find the risk factors for LT or death after MARS treatment (second analysis). Receiver operating characteristic (ROC) curves were performed on significant variables in the logistic regression model with the intent to individually predict variables for LT or death. After a stepwise multivariable logistic regression analysis enrolling all of the previously reported features only 2 variables, tumor necrosis factor (TFN)-α and Glasgow coma score (GCS) score, were statistically significant. TNF-α was an unique independent risk factor for retransplantation or death after MARS treatment (odds ratio [OR] 1.235; P =.013). Conversely, GCS score was protective against retransplantation or death (OR 0.150; P =.003). Starting from these assumptions, a predictive model was created using these 2 variables. On ROC analysis, the combined score showed an area under the curve greater than that of the 2 variables considered separately. Validating these results with a larger number of patients, we considered these 2 factors as subjective parameters to determine outcomes and the difference between PNF and DGF. © 2012 Elsevier Inc. All rights reserved.