Background. Acute kidney injury (AKI) is one of the common complications after living donor liver transplantation (LDLT) and is associated with increased mortality and morbidity. The prognostic nutritional index (PNI) has been used as a predictive model for postoperative complications. Here, we create a new predictive model based on the PNI and compared its predictive accuracy to other models in patients who underwent LDLT. Material and Methods: The data from 423 patients were collected retrospectively. The patients were dichotomized into the non-AKI and the AKI groups. Multivariate adjustment for significant postoperative variables based on univariate analysis was performed. A new predictive model was created using the results from logistic regression analysis, dubbed the modified-PNI model (mPNI). The area under the receiver operating characteristic curve (AUC) was generated to determine the diagnostic accuracy and cutoff value of individual models. The net reclassification improvement (NRI) and integrated discrimination improvement (IDI) were calculated to investigate diagnostic improvement by the mPNI. Results: Fifty-four patients (12.7 %) were diagnosed with AKI within 1-week after LDLT. The mPNI had the highest predictive accuracy (AUC = 0.823). The model of end-stage liver disease (MELD) scores and PNI were 0.793 and 0.749, respectively, and the INR and serum bilirubin were 0.705 and 0.637, respectively. The differences in the AUCs were statistically significant among the mPNI, PNI, INR, and serum bilirubin. The cutoff value for mPNI was 8.7. The NRI was 10.4% and the IDI was 3.3%. Conclusions: The mPNI predicted AKI within 1-week better than other scoring systems in patients who underwent LDLT. The recommended cutoff value of mPNI is 8.7.

No abstract