Liver re-transplantation: internal validation of a predictive mathematical model of survival

Because of the worse results from re-transplantation, a model for determining the long-term survival has been previously developed. Its effectiveness had to be tested and validated, as proposed in this study, using a different sample population than the one used to build it.

Age, recipient creatinine, urgency of re-transplantation, interval between primary liver transplant and re-transplantation (RETx) of 92 patients that received first liver RETx, from a different sample of patients, in a different time period than those used to develop the initial model. The proposed mathematical model was used to predict survival at six months after undergoing liver RETx. We compared the areas under the ROC curves (AROC) corresponding to the two independent samples (derivation and validation samples). By the log-rank technique, the survival curves were also compared and classified into tertiles according to the risk scores of the original model: high risk (>32), medium risk (24-32) and low risk (<24).

Age, creatinine, time between primary liver transplant and re-transplantation and the urgency with which patients were enrolled, had comparable survival curves among the derivation and validation samples. When comparing the AROC of the derivation (0.733) and validation (0.741) samples, there was no statistically significant difference (p=0.915), therefore sensitivity and specificity ratios between the two are similar.

This study made it possible to internally validate the original model for predicting survival at six months after undergoing liver RETx, although an external validation still needs to be done.